DEMOCRITUS UNIVERSITY OF THRACE
DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING
SOFTWARE AND APPLICATIONS DEVELOPMENT SECTOR
Towards a Future Internet Architecture
“Protocols and Support Mechanisms”
PhD Dissertation
Sotiris-Angelos Lenas
Xanthi, November 2015
ΔΗΜΟΚΡΙΤΕΙΟ ΠΑΝΕΠΙΣΤΗΜΙΟ ΘΡΑΚΗΣ
ΤΜΗΜΑ ΗΛΕΚΤΡΟΛΟΓΩΝ ΜΗΧΑΝΙΚΩΝ ΚΑΙ ΜΗΧΑΝΙΚΩΝ ΥΠΟΛΟΓΙΣΤΩΝ
ΤΟΜΕΑΣ ΛΟΓΙΣΜΙΚΟΥ ΚΑΙ ΑΝΑΠΤΥΞΗΣ ΕΦΑΡΜΟΓΩΝ
Αρχιτεκτονική του Μελλοντικού Διαδικτύου
“Πρωτόκολλα και Μηχανισμοί Υποστήριξης”
Διδακτορική Διατριβή
Σωτήριος – Άγγελος Λένας
Ξάνθη, Νοέμβριος 2015
Dedicated to my parents.
You made great sacrifices for me,
supported me endlessly and
loved me unconditionally.
Thank you.
ACKNOWLEDGEMENTS
Looking back on the past four years, I have come to realize that this journey would
not have been possible without the support of my family, professors, colleagues and
friends. I hereby take the opportunity to acknowledge the people who helped me
complete this thesis.
First of all, I would like to express my deepest gratitude to Professor Vassilis
Tsaoussidis, the advisor of my PhD thesis, for all the opportunities he gave me to
grow, both professionally and personally. At every step of this thesis, his valuable
guidance and sharp academic insights filled me with enthusiasm to explore new
ideas and provided me with motivation to keep moving forward. Above all, his
positive attitude, strong academic ethics and the trust that he shows to the team are
all characteristics of an exceptional mentor that I was fortunate to have as a PhD
student.
I would also like to thank the rest of my advisory committee members, Professor
Timos Sellis and Associate Professor Pavlos Efraimidis, for attending all necessary
academic meetings and dealing with all the administrative issues, as well as the
members of my dissertation committee, Professors Alexander Karakos, Avgerinos
Arampatzis, Ioannis Stavrakakis and Jon Crowcroft for the evaluation of my work.
My special thanks also go to Scott Burleigh for the constructive collaboration that
we had during my visit at NASA’s Jet Propulsion Laboratory. I am grateful to have
worked alongside such a gifted communication systems engineer and an
extraordinary person. Dr. Arjuna Sathiaseelan is another friend and colleague from
Cambridge University that I would like to thank for sharing the Public Access Wi-Fi
project traffic traces, which were employed in this thesis to evaluate the
performance of Hybrid Packet Scheduling Scheme.
Next, I would like to thank my colleagues at DUTH: Sotiris Diamantopoulos, Dimitris
Vardalis, Giorgos Papastergiou, Nikos Bezirgiannidis, Yannis Komnios, Fani Tsapeli,
i
ii
Acknowledgements
Stelios Dimitriou, Yannis Alexiadis, Christina Malliou and Argyris Samourkasidis for
their fruitful collaboration in our joint research efforts, all the great experiences that
we shared together and for creating an amazing work environment; Lefteris
Mamatas and Christos Samaras for frequently sharing their scientific ideas and
outlooks; and, Agapi Papakonstantinou for her skillful and caring administrative
support. It is a privilege for me to work and share my life with so many bright
people.
As we are getting closer to the end of this list of acknowledgements − a list that I
wish could go on forever − I would be remiss if I did not thank all my close friends
for their continuous encouragement and support throughout this challenging
period; I would never have gotten this far without you. So, thank you guys and gals
for all the laughs, fruitful discussions and moments that you shared with me. A
special mention has to go to my good friend Nana Koliousi for her time and effort in
editing a great part of this thesis.
Last, but certainly not least, I would like to thank my parents for their never
diminishing faith and the unwavering support they have offered me throughout my
life, despite their own personal battles. Dedicating this thesis to you guys is only but
a small token of my gratitude.
ABSTRACT
Internet's fundamental role in most aspects of modern life constitutes, beyond any
doubt, a fact. Its influence on today’s life is expected to be further increased in the
near future, given the interconnection of different types of devices, the support for
myriads of applications and the provision of ubiquitous network connectivity,
independently of location and user privileges; thus, signaling the dawn of the Future
Internet era. Nevertheless, various design and operational limitations of the current
Internet architecture restrict the establishment of a sufficiently broad networking
platform. Therefore, the Internet architecture, as it currently stands, is unable to
satisfy the properties of Global Architecture, Global Reach and Global Empowerment
that the Future Internet needs to exhibit. To this end, this thesis focuses on solutions
for a number of challenging issues related to the Future Internet.
In this context, Delay-/Disruption-Tolerant Networking (DTN) is proposed as a
network architecture that is suitable to support the formation of the Future Internet
by addressing most of current Internet architecture limitations in terms of fostering
interoperability among heterogeneous network environments. More specifically, it
is argued that DTN could play a key role in this formation process: first, by allowing
users and devices from heterogeneous networking environments to transparently
communicate with each other, independently of the underlying protocol stack, and,
second, by acting as a service layer for mitigating various network connectivity
issues, such as long delays and connectivity disruptions, that occur mainly due to
the continually extended use of wireless communications.
With a view to further enhance the operational capabilities of a DTN-based Future
Internet architecture, specific solutions for real-world applications are proposed
and developed, with a special emphasis placed upon data dissemination, availability,
and accessibility topics. In particular, in terms of data dissemination, Bundle
Streaming Service (BSS) is developed to enhance Future Internet data streaming
capabilities. BSS constitutes a DTN-based data streaming content distribution
iii
iv
Abstract
framework that can be employed to alleviate most of the networking challenges
related to both live and stored data streaming over heterogeneous network
environments. With respect to network availability and accessibility topics, a fully
distributed, scalable, Space-based, data dissemination system is proposed in order
to broaden Future Internet availability and widen the accessibility of its resources.
Through this system, data can be asynchronously delivered to interested end-users
across the globe; an approach that significantly increases the cost-effectiveness of
the proposed system — thus, making it affordable and applicable to a much wider
audience. Furthermore, Hybrid Packet Scheduling Scheme (HPSS) is developed to
widen the accessibility on Future Internet network resources by enhancing the
efficiency of broadband sharing schemes. HPSS constitutes a queue-management
framework that allows for a more effective exploitation of the available broadband
network resources by guest users; thus, allowing Internet access to further broaden
without any additional cost.
Extensive analytical, simulation and emulation studies are performed, accordingly,
to assess both the efficiency and the effectiveness of the aforementioned solutions.
The respective evaluation results provide useful insights into the dynamics of the
various Future Internet topics explored herein.
In conclusion, this thesis not only contributes to the theoretical perspective of
Future Internet deployment by exploring the networking challenges and
opportunities stemming from changes in current Internet architecture but also
sheds light on many real-world applications. In this regard, it shows that: i) it is
possible to efficiently transfer data streams over heterogeneous internetworked
environments, ii) the incorporation of Space assets into a Future Internet
architecture and the implementation of an asynchronous data delivery system is
technically feasible by utilizing already-existing and proven-to-work networking
technologies,
iii)
unused
bandwidth
capacity
from
broadband
network
infrastructures can be exploited in a fully transparent and scalable manner to
strengthen the overall global networking participation and enhance the social
impact of the Future Internet.
ΠΕΡΙΛΗΨΗ
Ο θεμελιώδης ρόλος του Διαδικτύου στις περισσότερες πτυχές του σύγχρονου
τρόπου ζωής αποτελεί αδιαμφισβήτητο γεγονός. Η επιρροή που ασκεί το Διαδίκτυο
σήμερα αναμένεται να αυξηθεί περαιτέρω στο προσεχές μέλλον με τη διασύνδεση
συσκευών διαφορετικού τύπου, την υποστήριξη πληθώρας εφαρμογών και την
παροχή καθολικής δικτυακής συνδεσιμότητας, ανεξαρτήτως τοποθεσίας και
προνομιών χρήστη σηματοδοτώντας με τον τρόπο αυτό την απαρχή της εποχής του
Μελλοντικού Διαδικτύου. Παρόλα αυτά, διάφοροι σχεδιαστικοί και λειτουργικοί
περιορισμοί της υπάρχουσας αρχιτεκτονικής του Διαδικτύου εμποδίζουν την
εγκαθίδρυση μιας τόσο διευρυμένης διαδικτυακής πλατφόρμας. Επομένως, η
αρχιτεκτονική του Διαδικτύου, στην παρούσα μορφή της, είναι ανίκανη να
υποστηρίξει τις συνθήκες της Καθολικής Αρχιτεκτονικής, Καθολικής Διαθεσιμότητας
και Καθολικής Πρόσβασης τις οποίες το Μελλοντικό Διαδίκτυο θα πρέπει να
ικανοποιεί. Έχοντας υπόψη αυτούς του περιορισμούς, η διατριβή επικεντρώνεται
στην πρόταση και την ανάπτυξη λύσεων για μια πληθώρα απαιτητικών ζητημάτων
που σχετίζονται με το Μελλοντικό Διαδίκτυο.
Σε αυτό το πλαίσιο, προτείνεται η χρήση της αρχιτεκτονικής δικτύων ανεκτικών
στην καθυστέρηση και στις διακοπές (Delay-/Disruption-Tolerant Networking -
DTN) ως μιας διαδικτυακής αρχιτεκτονικής που δύναται να συνεισφέρει
καθοριστικά στον σχηματισμό του Μελλοντικού Διαδικτύου. Η κύρια συνεισφορά
της έγκειται στο γεγονός ότι μέσω των μηχανισμών που προσφέρει καθιστά
δυνατή την αντιμετώπιση των αδυναμιών της υπάρχουσας αρχιτεκτονικής του
Διαδικτύου, κυρίως ως προς την διασύνδεση ετερογενών διαδικτύων. Πιο
συγκεκριμένα, υποστηρίζεται πως η αρχιτεκτονική DTN μπορεί να διαδραματίσει
κεντρικό ρόλο στην διαδικασία σχηματισμού του Μελλοντικού Διαδικτύου: αρχικά,
επιτρέποντας σε χρήστες και συσκευές ετερογενών δικτυακών περιβαλλόντων να
επικοινωνούν μεταξύ τους, ανεξαρτήτως της υποκείμενης στοίβας πρωτοκόλλων
και εν συνεχεία, δρώντας ως ένα δικτυακό στρώμα μέσω του οποίου θα είναι
v
vi
Περίληψη
δυνατή η επίλυση διαφόρων ζητημάτων συνδεσιμότητας, όπως, για παράδειγμα, οι
μεγάλες καθυστερήσεις και οι συνεχείς διακοπές επικοινωνίας, τα οποία
προκύπτουν κυρίως από την αυξανόμενη χρήση ασύρματων τεχνολογιών
επικοινωνίας.
Με σκοπό την περαιτέρω ενίσχυση των λειτουργικών δυνατοτήτων μιας
αρχιτεκτονικής του Μελλοντικού Διαδικτύου η οποία θα βασίζεται στην
αρχιτεκτονική DTN, προτείνονται και υλοποιούνται επίσης δικτυακές λύσεις για
πρακτικές
εφαρμογές,
δίνοντας
ιδιαίτερη
έμφαση
σε
θέματα
διάχυσης,
διαθεσιμότητας και προσβασιμότητας δεδομένων. Συγκεκριμένα, σχετικά με το
θέμα της διάχυσης δεδομένων, υλοποιείται το Bundle Streaming Service (BSS) με
σκοπό την ενίσχυση των δυνατοτήτων του Μελλοντικού Διαδικτύου ως προς τη
μεταφορά ροών δεδομένων. Το BSS αποτελεί πλαίσιο διαχείρισης ροών δεδομένων,
το οποίο βασίζεται στην αρχιτεκτονική DTN και μπορεί να χρησιμοποιηθεί για την
άμβλυνση των περισσότερων δικτυακών προκλήσεων που σχετίζονται με την
μεταφορά ροών δεδομένων, τόσο πραγματικού όσο και μεταγενέστερου χρόνου,
μεταξύ ετερογενών διαδικτύων. Όσον αφορά τα ζητήματα της προσβασιμότητας
και της διαθεσιμότητας, προτείνεται ένα πλήρως κατανεμημένο, επεκτάσιμο και
βασιζόμενο σε δορυφορικές επικοινωνίες σύστημα διάχυσης δεδομένων. Μέσω
αυτού του συστήματος, δεδομένα μπορούν να μεταφερθούν με ασύγχρονο τρόπο
σε ενδιαφερόμενους τελικούς χρήστες σε κάθε σημείο της γης, επεκτείνοντας με
αυτό τον τρόπο την διαθεσιμότητα του Μελλοντικού Διαδικτύου. Η συγκεκριμένη
δυνατότητα της ασύγχρονης μεταφοράς δεδομένων αυξάνει σημαντικά την
αποδοτικότητα ως προς το κόστος του προτεινόμενου συστήματος, καθιστώντας
το προσιτό και προσβάσιμο από το ευρύ κοινό. Τέλος, υλοποιείται ο μηχανισμός
Hybrid Packet Scheduling Scheme (HPSS) με σκοπό την περαιτέρω ενίσχυση της
προσβασιμότητας στους δικτυακούς πόρους του Μελλοντικού Διαδικτύου, μέσω της
ενίσχυσης της αποδοτικότητας των συστημάτων διαμοιρασμού ευρυζωνικής
σύνδεσης. Ο μηχανισμός HPSS αποτελεί ένα πλαίσιο διαχείρισης ουράς μέσω του
οποίου επιτυγχάνεται αποδοτικότερη εκμετάλλευση των διαθέσιμων πόρων μια
Αρχιτεκτονική του Μελλοντικού Διαδικτύου: Πρωτόκολλα και Μηχανισμοί vii
Υποστήριξης
ευρυζωνικής σύνδεσης από επισκέπτες χρήστες επιτρέποντας με αυτό τον τρόπο
την επέκταση του Μελλοντικού Διαδικτύου, χωρίς κανένα επιπλέον κόστος.
Εκτενείς μελέτες ανάλυσης, προσομοιώσεις και εξομοιώσεις υλοποιούνται, ανάλογα
με την εκάστοτε περίπτωση, ώστε να αξιολογηθεί τόσο η αποδοτικότητα όσο και η
αποτελεσματικότητα των προαναφερθέντων δικτυακών λύσεων. Τα αποτελέσματα
της αξιολόγησης παρέχουν πληθώρα χρήσιμων πληροφοριών σχετικά με τη
δυναμική των διαφόρων θεμάτων που διερευνώνται στα πλαίσια της διατριβής και
σχετίζονται με το Μελλοντικό Διαδίκτυο.
Συμπερασματικά, η συνεισφορά της διατριβής δεν εξαντλείται στη θεωρητική
μελέτη σχηματισμού του Μελλοντικού Διαδικτύου, μέσω της διερεύνησης των
δικτυακών προκλήσεων που προκύπτουν από αλλαγές στην υπάρχουσα
αρχιτεκτονική του Διαδικτύου, αλλά επεκτείνεται και σε ζητήματα που αφορούν
πρακτικές εφαρμογές. Από αυτή την σκοπιά, συνάγεται μέσω των αποτελεσμάτων
που παρουσιάζονται ότι: α) είναι δυνατή η αποδοτική μεταφορά ροών δεδομένων
μεταξύ ετερογενών δικτύων, β) η ενσωμάτωση δορυφορικών στοιχείων
επικοινωνίας στην αρχιτεκτονική του Μελλοντικού Διαδικτύου καθώς και η
υλοποίηση ενός συστήματος ασύγχρονης διάχυσης δεδομένων αποτελούν τεχνικά
εφικτές ενέργειες στηριζόμενες σε υπάρχουσες και ενδελεχώς δοκιμασμένες
δικτυακές τεχνολογίες και γ) οι διαθέσιμοι πόροι υποδομών διαμοιρασμού
ευρυζωνικής σύνδεσης μπορούν να αξιοποιηθούν με έναν πλήρως διαφανές, ως
προς τον τελικό χρήστη, και επεκτάσιμο τρόπο, ώστε να ενισχυθεί συνολικά η
προσβασιμότητα στο Μελλοντικό Διαδίκτυο, ενδυναμώνοντας περαιτέρω κατά
αυτόν τον τρόπο την απήχηση του στην κοινωνία.
viii
Περίληψη
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ............................................................................................................ I
ABSTRACT ................................................................................................................................. III
TABLE OF FIGURES ................................................................................................................. XI
LIST OF TABLES .................................................................................................................... XIII
LIST OF ABBREVIATIONS .................................................................................................... XV
1. INTRODUCTION ................................................................................................................. 21
1.1. THESIS DESCRIPTION ...................................................................................................................... 21
1.2. MOTIVATION .................................................................................................................................... 21
1.2.1. Global Architecture ...............................................................................................................23
1.2.2. Global Reach .............................................................................................................................23
1.2.3. Global Empowerment ...........................................................................................................24
1.3. AIMS AND OBJECTIVES .................................................................................................................... 25
1.4. CONTRIBUTION AND MAIN CONCLUSIONS ................................................................................... 26
1.5. IMPACT .............................................................................................................................................. 29
1.6. OUTLINE............................................................................................................................................ 30
2. DELAY-/DISRUPTION-TOLERANT NETWORKING AS A FUTURE GLOBAL
INTERNETWORKING ARCHITECTURE ............................................................................ 33
2.1. CURRENT INTERNET ARCHITECTURE LIMITATIONS................................................................... 33
2.2. TO “CLEAN-SLATE” OR NOT TO “CLEAN-SLATE”? ...................................................................... 35
2.3. DELAY-/DISRUPTION-TOLERANT NETWORKING....................................................................... 37
2.3.1. Key Architectural Principles (Pros) ................................................................................38
2.3.2. Architectural Shortcomings (Cons) ................................................................................44
2.4. TOWARDS A DTN-BASED FUTURE INTERNET ARCHITECTURE ................................................ 46
3. ENHANCING DATA STREAMING FUNCTIONALITY IN DELAY-/ DISRUPTIONTOLERANT NETWORKS ....................................................................................................... 49
3.1. BACKGROUND................................................................................................................................... 50
3.2. BUNDLE STREAMING SERVICE ....................................................................................................... 53
3.2.1. Design and Implementation Details ...............................................................................54
3.2.2. Forwarding Algorithm .........................................................................................................57
3.2.3. Display Algorithm ..................................................................................................................58
4. EXTENDING INFORMATION ACCESS THROUGH AN ASYNCHRONOUS, SPACEORIENTED DATA DISSEMINATION SYSTEM ................................................................. 61
4.1. BACKGROUND AND DESIGN GUIDELINES...................................................................................... 62
4.2. A FULLY DISTRIBUTED AND ASYNCHRONOUS DATA DISSEMINATION MODEL ...................... 66
4.3. A DTN-BASED SPACE-TO-GROUND DATA DISSEMINATION SYSTEM ...................................... 68
4.3.1. Data Fragmentation .............................................................................................................69
4.3.2. Addressing .................................................................................................................................70
ix
x
Table of Contents
4.3.3. Routing .......................................................................................................................................70
4.3.4. Reliability ...................................................................................................................................70
5. WIDENING INTERNET ACCESSIBILITY THROUGH BROADBAND SHARING ... 73
5.1. BACKGROUND ................................................................................................................................... 75
5.2. BROADBAND SHARING SYSTEM MODELING................................................................................. 77
5.2.1. Numerical Analysis ................................................................................................................78
5.3. TOWARDS A HYBRID PACKET SCHEDULING SCHEME ................................................................ 82
5.3.1. Impact of Different Packet-size Distributions and Bandwidth Capacities on
Average Global System Time .........................................................................................................82
5.3.2. Impact of Guest-user Traffic on Home-user Traffic .................................................83
5.4. HYBRID PACKET SCHEDULING SCHEME ....................................................................................... 86
6. EVALUATION METHODOLOGY ...................................................................................... 89
6.1. EVALUATION OBJECTIVES............................................................................................................... 89
6.2. DATASETS ANALYSIS FOR BROADBAND SHARING SIMULATIONS ............................................. 92
6.2.1. Modeling Access Points ........................................................................................................93
6.2.2. Modeling Guest-user Traffic ..............................................................................................95
6.3. SCENARIOS ..................................................................................................................................... 102
6.3.1. Experimental Category I – Data Streaming............................................................. 102
6.3.2. Experimental Category II – Asynchronous, LEO-based, Data Dissemination
Model
............................................................................................................................................. 108
6.3.3. Experimental Category III – Broadband Sharing .................................................. 111
6.4. METRICS ........................................................................................................................................ 115
6.5. EXPERIMENTATION TOOLS ......................................................................................................... 116
6.5.1. Data Streaming Emulations ........................................................................................... 116
6.5.2. Asynchronous, LEO-based, Data Dissemination Model Simulations ............. 120
6.5.3. Broadband Sharing Simulations................................................................................... 125
7. EVALUATION RESULTS .................................................................................................. 127
7.1. EXPERIMENTAL CATEGORY I – DATA STREAMING .................................................................. 127
7.1.1. Network Performance ....................................................................................................... 127
7.1.2. Multimedia Performance ................................................................................................. 131
7.2. EXPERIMENTAL CATEGORY II – ASYNCHRONOUS, LEO-BASED, DATA DISSEMINATION
MODEL .................................................................................................................................................. 132
7.3. EXPERIMENTAL CATEGORY III – BROADBAND SHARING ....................................................... 134
8. CONCLUSIONS AND OPEN ISSUES .............................................................................. 141
8.1. BUNDLE STREAMING SERVICE .................................................................................................... 144
8.2. ASYNCHRONOUS, LEO-BASED, FULLY-DISTRIBUTED DATA DISSEMINATION SYSTEM ...... 144
8.3. HYBRID PACKET SCHEDULING SCHEME .................................................................................... 145
REFERENCES .......................................................................................................................... 149
TABLE OF FIGURES
Figure 1-1. Cisco’s estimation on Internet-connected devices by 2020 (Source: [2]). ....22
Figure 3-1. Bundle Streaming Service architecture. .....................................................................56
Figure 4-1. Data return model with multiple end-users receiving data through
broadcasting...................................................................................................................................................66
Figure 4-2. DTN deployment for implementing a distributed data return model. ..........68
Figure 4-3. Abstract DTN-based protocol stack for each network entity. ...........................69
Figure 5-1. Average system time for various packet-size distributions and bandwidth
capacities. ........................................................................................................................................................83
Figure 5-2. Average system time for both classes under several combinations of
channel utilization. ......................................................................................................................................83
Figure 5-3. Percentage impact of LBE-class traffic over BE-class traffic. ...........................85
Figure 5-4. Temporal impact of LBE-class traffic over BE-class traffic. ...............................85
Figure 5-5. Temporal impact of LBE-class traffic over BE-class traffic on high traffic
loads. ..................................................................................................................................................................86
Figure 5-6. Additional delay imposed to the average system time of BE-class traffic. ...87
Figure 6-1. Public Access Wi-Fi architecture....................................................................................93
Figure 6-2. Access network profiles produced from the analysis of data collected
during the PAWS project trial.................................................................................................................94
Figure 6-3. P-P plots comparing the empirical distribution of inter-arrival time, flow
duration and flow size characteristics with various statistical distributions. ...................99
Figure 6-4. All three network topologies employed in BSS-specific Space scenarios. .. 103
Figure 6-5. All three network properties classes of the topology employed in BSSspecific terrestrial scenarios................................................................................................................. 105
xi
xii
Table of Figures
Figure 6-6. BER and propagation delay distribution on each 5-node network topology
variation.. ...................................................................................................................................................... 107
Figure 6-7. Content distribution scenario network topology. ................................................ 108
Figure 6-8. Network topology considered for broadband sharing simulations. ............ 111
Figure 6-9. SPICE DTN testbed architecture. ................................................................................ 117
Figure 6-10. Representation of the network stack used in both configurations. ........... 119
Figure 6-11. Transmission buffer example with 4 ADUs of 5 PDUs each. ......................... 125
Figure 7-1. Comparison between BSSP and IPN forwarder for the a) 3-node and b) 5node network topology of all three terrestrial scenario variations. ................................... 128
Figure 7-2. Comparison between BSSP and IPN forwarder for the a) 3-node and b) 5node network topology of all three Space scenarios. ................................................................. 130
Figure 7-3. Comparison between BSSP and IPN forwarder based on SDT metric of a
representative sample of cases from both a) terrestrial and b) Space scenarios. ......... 131
Figure 7-4. Comparison between BSSP and IPN forwarder based on a) EDE b) EDA and
c) PSNR metrics for the 3-node network topology of the Moon Space scenario. ........... 131
Figure 7-5. Video stream snapshots of a) BSSP and b) IPN forwarder for the 3-node
network topology of the Moon Space scenario. ............................................................................ 132
Figure 7-6. Low-cost broadcast delivery ratio and delivery latency varying the number
of ground stations. .................................................................................................................................... 133
Figure 7-7. Low-cost broadcast delivery ratio and data volume varying the number of
satellites......................................................................................................................................................... 133
Figure 7-8. Broadband sharing simulation results. .................................................................... 136
LIST OF TABLES
Table 5-1. List of the Examined Packet-size Distributions. ........................................................79
Table 5-2. Notation Table. ........................................................................................................................79
Table 6-1. User Profile Distribution Parameters. ........................................................................ 101
Table 6-2. Average Fitting Error Between Empirical and Analytical Model. .................. 102
Table 6-3. Contact Opportunities Statistics. .................................................................................. 110
xiii
xiv
List of Tables
LIST OF ABBREVIATIONS
Abbreviation/Term
Definition
ADSL
Asymmetric Digital Subscriber Line
AOS
Advanced Orbiting Systems
BE
Best Effort
ADU
Application Data Unit
BER
BP
Bit Error Rate
Bundle Protocol
BRAS
Broadband Remote Access Server
BSSP
Bundle Streaming Service Protocol
BxD
Bandwidth-Delay Product
CBR
Constant Bit Rate
CCSDS
Consultative Committee for Space Data Systems
BSS
BW
CB
Bundle Streaming Service
Bandwidth
Class-based
xv
xvi
List of Abbreviations
CFDP
CODEL
CSMA/CA
DMC
DSCP
CCSDS File Delivery Protocol
Controlled Delay
Carrier Sense Multiple Access with Collision
Avoidance
Disaster Monitoring Constellation
Differentiated Services Code Point
DTN
Delay-/Disruption-Tolerant Networking
DVB-S
Digital Video Broadcasting – Satellite
DTNs
E2E
Delay-/Disruption-Tolerant Networks
End-to-End
EDE
End-user’s Display Efficiency
ETSI
European Telecommunications Standards Institute
FCFS
First Come First Served
EID
EU
FEC
FI
Endpoint Identifier
European Union
Forward Error Correction
Future Internet
Towards a Future Internet Architecture: Protocols and Support Mechanisms xvii
FIArch
Future Internet Architecture
FIFO
First In First Out
FIF
Future Internet Forum
FIND
Future Internet Design
FIRE
Future Internet Research and Experimentation
FQ
FTP
GENSO
Fair Queueing
File Transfer Protocol
Global Educational Network for Satellite
Operations
GENI
Global Environment for Network Innovations
GS
Ground Station
GEO
Geosynchronous Equatorial Orbit
HPSS
Hybrid Packet Scheduling Scheme
ION
Interplanetary Overlay Network
IP
Internet Protocol
HTTP
IoT
Hypertext Transfer Protocol
Internet of Things
xviii
List of Abbreviations
IPN
InterPlanetary Networking
ITU
International Telecommunication Union
LAN
Local Area Network
ISP
LBE
Internet Service Provider
Less than Best Effort
LEDBAT
Low Extra Delay Background Transport
LTP
Licklider Transmission Protocol
MANET
Mobile Ad-hoc Network
MEO
Medium Earth Orbit
LEO
MCC
Low Earth Orbit
Mission Control Center
MRG
Minimum Reception Group
NDN
Named Data Networking
PAWS
Public Access Wi-Fi Service
PDU
Protocol Data Unit
Towards a Future Internet Architecture: Protocols and Support Mechanisms xix
PER
Packet Error Rate
PI
Principal Investigator
PSNR
Peak Signal-to-Noise Ratio
RED
Random Early Detection
PQ
Priority Queueing
RTS/CTS
Request-To-Send/Clear-To-Send
SDA
Stream Delivery Attenuation
SDT
Stream Delivery Time
RTT
SDE
Round-Trip Time
Stream Delivery Efficiency
SRR
Smoothed Round Robin
TCP
Transmission Control Protocol
TDRS
Tracking and Data Relay Satellite
ToS
Type-of-Service
TCPCL
TM/TC
TCP Convergence Layer
Telemetry/Telecommand
xx
List of Abbreviations
TTL
Time-to-Live
UDP
User Datagram Protocol
UDPCL
UDP Convergence Layer
URI
Uniform Resource Identifier
WFQ
Weighted Fair Queueing
UPN
USA
WLAN
WNIC
WSN
User Provided Network
United States of America
Wireless Local Area Network
Wireless Network Interface Card
Wireless Sensor Network
1. Introduction
1.1. Thesis Description
Internet's fundamental role in most aspects of modern life constitutes, beyond any
doubt, a fact. Its influence on today’s life is expected to be further increased in the
near future, given the interconnection of different types of devices, the support for
myriads of applications and the provision of ubiquitous network connectivity,
independently of location and user privileges; thus, signaling the dawn of the Future
Internet era. Nevertheless, various design and operational limitations of the current
Internet architecture restrict the establishment of a sufficiently broad networking
platform. Within this context, this thesis focuses on solutions for a number of
challenging issues related to the Future Internet: it proposes and develops
communication models, protocols and support mechanisms that emphasize on
data dissemination, availability and accessibility issues.
1.2. Motivation
The Internet has become an indispensable part of our everyday life. Although it
started as an academic network, after an evolutionary process that took a few
decades, it expanded to become the broad informational, commercial and
communication platform that it is today. Despite its tremendous growth, Internet’s
expansion still seems to be at an early stage. Currently, the Internet is available to
21
22 Chapter 1: Introduction
only 39% of the world population [1], with almost 4 billion people left without
Internet access. Furthermore, recent studies suggest that in the near future, even
without taking into account any networking or device-related advances, the number
of Internet-connected devices will be almost doubled [2]. In particular, as Figure 1-1
shows, according to CISCO’s estimations, almost 50 billion devices will be connected
to the Internet by 2020, with Internet of Things (IoT) as the main driving force
behind this explosion.
Figure 1-1. Cisco’s estimation on Internet-connected devices by 2020 (Source: [2]).
Internet’s expansion, though, is not likely to stop there. New networking trends and
communication paradigms are constantly emerging due to various socio-economic,
business and environmental needs, forcing the Internet to transform into an even
broader access platform than it is today. By taking into account this continuous
expansion and irrespectively of any particular future developments, the Future
Internet, as this broader access platform is named, should be able to provide all
kinds of devices with the ability to access the Internet, anytime, anywhere. As S.
Burleigh et al. suggested in a recent work regarding the establishment of a future
global Internet service [3], in order for such a broader access platform as the Future
Internet to be fully realized, it needs to exhibit three major properties: i) Global
Architecture, ii) Global Reach and iii) Global Empowerment. These three properties
constitute the motivational pillar on which the work presented in this thesis rests.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 23
1.2.1. Global Architecture
Global architecture refers to the seamless integration of heterogeneous networking
environments. Apart from the Internet and cellular networks, Future Internet is
expected to incorporate several types of emerging networking environments,
including Mobile Ad-hoc Networks (MANETs), Vehicular Ad-hoc Networks, Wireless
Sensor Networks (WSNs), Wireless Personal Area Networks, Wireless Body Area
Networks, Underwater Wireless Sensor Networks, Deep Space Networks,
Interplanetary Networks (IPNs), Low Earth Orbiting satellite networks etc. This
highly heterogeneous environment must behave as a fully integrated and
interconnected network, in which either fixed or mobile users can communicate
with each other independently of the underlying network technology. Therefore,
this new internetwork, i.e. network of networks, must inherently support nodes’
mobility along with multiple forms of network heterogeneity, including diverse
communication mediums (e.g. fiber, wireless, satellite communications) with
variable bandwidths, end-to-end delays and connection availability periods, various
local routing and management mechanisms, and, finally, numerous protocol stacks.
All these diverse requirements, along with the fact that maintaining an end-to-end
connection between two communicating hosts cannot be guaranteed in such a
complex, non-deterministic networking environment, render the TCP/IP Internet
architecture inappropriate for solely supporting the Future Internet. Thus, another,
more distributed network architecture should be selected that provides the
necessary support and functionalities for seamlessly integrating all those
heterogeneous networking environments into a unified networking framework.
1.2.2. Global Reach
Global reach refers to information access that is available wherever information may
be needed or produced. In this respect, terrestrial communication infrastructures
are inevitably inadequate, since their deployment is geographically limited. Satellite-
based communication systems, however, can beam data to any point of the Earth’s
surface with equal ease.
24 Chapter 1: Introduction
Currently, most satellite-based Internet services are relying on geosynchronous
(GEO) satellite communication systems for their operation. Those systems are
essentially acting as analogs for wired infrastructure, enabling continuous end-to-
end interaction between communicating entities. This particular approach presents
however three significant drawbacks: i) GEO satellites are providing partial
coverage of the Earth’s surface, ii) in order for an organization to provide a global
Internet broadband service, it has to deploy several GEO satellites, which constitutes
a rather costly process and iii) due to their high altitude, GEO-based satellite
communication systems suffer from long propagation delays.
In an effort to compensate for these drawbacks, several Low-Earth-Orbit (LEO)
satellite communication systems, composed of tens of satellites, have been deployed
during the last decade [4]. Although those systems indeed look promising in
broadening Internet access to a global scale, their design philosophy of simply acting
as analogs for wired infrastructure by building end-to-end connections, in
combination with their deployment philosophy of extending users’ paid services,
confines their role as broad access platforms.
1.2.3. Global Empowerment
Global empowerment refers to information access that is available to everyone who
may need it to sustain enriched human life. Terrestrial network infrastructures are
subject to communication constraints and privacy infringement practices, such as
Internet censorship, digital surveillance and information exchange blockage within
national borders. Oppressive governments usually apply such practices in their
effort to prohibit open access to knowledge and information, as it may sometimes
seem to be in a government’s interest to limit its citizens’ access to information.
Recent examples of such information-constrain practices include the temporary
access restriction on several social networking platforms, such as Twitter and
YouTube, imposed by the Turkish government over the images of an Istanbul
prosecutor held hostage, and South Korea’s ongoing Internet censorship policy [5].
Satellite-based network infrastructure, however, is not subject to such practices,
since Space systems can provide global coverage. Although a satellite might be
Towards a Future Internet Architecture: Protocols and Support Mechanisms 25
under the control of a certain agency, company or nation, nothing prevents other
satellites, which serve different interests, from taking its place in the sky. Thus, a
Space-assisted network service could serve as a platform for helping oppressed
people achieve their civil and human rights globally.
Apart from the censorship and privacy infringement practices, economic-related
barriers are also among the main reasons that prohibit global access to information.
Fixed terrestrial networking infrastructures are absent from large parts of
underdeveloped countries, since no operator would proceed in such an investment
without the prospect of any profit in return. Space-assisted communication systems
could also be employed in this case to extend information access to a much wider
audience by providing free access to pre-loaded information via broadcasting or
even the Internet itself.
Even in developed countries, where terrestrial infrastructure is available, some part
of the population might be excluded from regularly accessing the Internet due to its
inability to afford some form of Internet service subscription fee. In such cases, the
exploitation of existing resource pooling Internet technologies could be employed to
reduce the negative consequences of digital exclusion faced by less-privileged
people by providing free Internet access to them.
1.3. Aims and Objectives
In light of the current Internet architecture limitations, as specifically identified in
the motivation section, and by taking into consideration, as well, the three
aforementioned fundamental Future Internet properties, this thesis aims to address
the following questions:
1) Which network architecture could serve as the basis for interconnecting
heterogeneous networking environments into a unified Future Internet
architecture?
2) Is it possible to efficiently transfer data streams over heterogeneous
internetworked environments? How such a universal data-streaming
framework could be implemented?
26 Chapter 1: Introduction
3) Is it possible for Space assets to be seamlessly incorporated into the Future
Internet architecture, in order to broaden its availability and widen the
accessibility of its resources? Could this be achieved in a fully distributed and
scalable manner?
4) Is it possible for unused bandwidth capacity from current network
infrastructures to be exploited, in order to strengthen the overall global
networking participation and enhance the social impact of the Future
Internet? Could this practice be applied to a massive scale and be
implemented in such a way that it is fully transparent to the end-user?
1.4. Contribution and Main Conclusions
The research effort expended during the preparation of this thesis has resulted in
the development of specialized solutions for: i) enhancing Future Internet data
dissemination capabilities through data streaming content distribution mechanisms,
ii) broadening Future Internet availability via satellite-based data dissemination
models, and iii) widening the accessibility on Future Internet network resources by
taking advantage of resource pooling techniques. Both the effectiveness and the
efficiency of these solutions have been extensively evaluated through realistic
networking scenarios, originated by representative terrestrial and Space real-world
use cases. Overall, the contribution of this thesis, along with the main conclusions,
can be summarized per Future Internet property as follows:
In terms of Global Architecture, it is argued that Delay-/Disruption-Tolerant
Networking (DTN) architecture [6] could play a key role in the formation of the
Future Internet by enabling the integration of heterogeneous networking
environments into a unified overlay network; thus, allowing users and devices from
different networking environments to transparently communicate with each other,
independently of the underlying protocol stack.
Nevertheless, in spite of the high-maturity level of both the DTN architecture and
the Bundle protocol (BP) [7], which is the main protocol that provides all the
necessary overlay network functionalities in Delay-/Disruption-Tolerant Networks
Towards a Future Internet Architecture: Protocols and Support Mechanisms 27
(DTNs), presently, there is a lack of standard mechanisms for efficiently supporting
data stream transfers over DTNs. This could be a limiting factor for Future Internet
data dissemination capabilities, especially when considering the fact that data
streaming constitutes one of the most popular content distribution methods
currently employed on the Internet. In order to compensate for this shortcoming,
Bundle Streaming Service (BSS) [8] is proposed and implemented, as a practical
framework through which many of the networking challenges related to both live
and stored data streaming over DTNs can be addressed.
The evaluation results regarding BSS show that it manages to enhance data
streaming delivery performance both in Space and terrestrial communications, by
allowing for a smoother near real-time streaming viewing experience over a variety
of data stream types, including multimedia streams. This is mainly attributed to the
sophisticated forwarding mechanism that BSS employs, with which different parts
of a data stream can be simultaneously forwarded through a combination of reliable
and unreliable transport services, based on the quality of the communication link.
In terms of Global Reach, an asynchronous, fully distributed, LEO-based data
dissemination model, whose implementation can be based on low-cost Space and
ground assets, is proposed that extends traditional satellite communications in
order to enable asynchronous delivery of data to interested end-users across the
globe [9].
Through the proposed model, data can be acquired from possibly disconnected end-
users, stored in satellite’s memory and physically transported over orbital tracks
until delivered to their destinations. The last step of this process could be
accomplished either by dumping the data on ground-segment gateways, which can
be transferred afterwards to the end-users’ via high-speed terrestrial links, or by
directly broadcasting the data to end-users’ terminals. In this way, both Future
Internet availability and accessibility are broadened significantly, since people in
remote areas lacking any terrestrial Internet infrastructure can have low cost access
to valuable data and information.
28 Chapter 1: Introduction
In order to investigate the effectiveness and the scalability of the proposed model,
its performance was quantified under different network setups and compared to
traditional satellite communication approaches. As expected, the performance gap
in terms of data return between the traditional approaches and the proposed model
decreases, as the number of low-cost satellites supporting the operation of the
proposed model increases, rendering the deployment of a low-cost broadcasting
distributed satellite communication system a promising alternative in economic
terms.
In terms of Global Empowerment, a mechanism for enabling free Internet access
was developed, as a means to widen Future Internet accessibility, extend its
applicability and enhance its social impact. In particular, the research effort was
focused on broadband sharing, a common resource pooling technique, which, apart
from widening Internet accessibility, is also associated with multiple other benefits,
such as reducing the negative consequences of digital exclusion faced by less-
privileged people, supporting emergency services, providing ubiquitous networking
and benefiting economic growth.
The implementation of broadband sharing systems is usually based on User
Provided Networks (UPNs). The concept of UPNs comprises two main entities: i)
home users, who act as micro-providers for guest users by building an access
network and sharing their Internet connection and ii) guest users, who wish to freely
access the Internet by exploiting the available unused capacity.
In this context, after numerically assessing the capability of broadband sharing
systems in providing an adequate level of service to guest users, Hybrid Packet
Scheduling Scheme (HPSS) [10] was developed, as a queue-management framework
that allows for a more effective exploitation of the available network resources by
guest users. To serve this purpose, HPSS dynamically decides what is the optimal
amount of resources that can be allocated for serving guest-user traffic at any time,
without impacting home users’ network performance; thus, allowing Internet access
to broaden further without any additional cost. In order to accomplish that, HPSS
identifies first the amount of available network resources, i.e. network resources
Towards a Future Internet Architecture: Protocols and Support Mechanisms 29
that are not exploited by home users, and then, based on resources availability, it
employs either a Priority Queuing (PQ) policy to protect home users from
experiencing periods of poor network performance, or a Weighted Fair Queuing
policy (WFQ) in order to guarantee a certain level of service for guest users.
After investigating the effectiveness of different broadband sharing scheme
configurations under various broadband connection setups, the results show that, in
general, domestic broadband connection sharing schemes could be used as a
framework for providing simultaneous free Internet access to a considerable
number of guest users. In comparison to other queueing methods, HPSS exhibited
the most balanced behavior by serving a considerable portion of guest-user traffic
without degrading home users’ network performance.
1.5.
Impact
The contribution of this thesis is multifold and spans across a number of network
research and engineering areas. Five particular directions are identified here:
1. It offers a theoretical perspective on the development of Future Internet by
exploring the networking challenges and opportunities stemming from
changes in current Internet architecture.
2. It extends the functional capabilities of DTN architecture and broadband
sharing infrastructures by implementing specifically-tailored solutions for
enhancing
data
streaming
transfers
over
heterogeneous
network
environments and improving exploitation of available network resources
from guest users.
3. It supports future low-cost Space network deployments by designing
asynchronous, Space-oriented data dissemination systems and by providing
a robust evaluation framework for assessing deployment plans.
4. It complements the existing literature by providing unique insights into the
dynamics of low-cost Space-based data dissemination systems and
broadband sharing infrastructures.
30 Chapter 1: Introduction
5. It assists the evolution towards a ubiquitous Future Internet service by
contributing to data dissemination, availability and accessibility aspects, as
described in the previous section.
Overall, the protocols and mechanisms introduced in this thesis propel Future
Internet’s development while, in the long term, they might also provide multiple
societal gains in terms of reducing the negative consequences of digital exclusion
faced by less-privileged people, restricting privacy infringement practices,
supporting emergency services, and, finally, benefiting economic growth.
1.6. Outline
The remainder of this thesis is organized as follows:
Chapter 2 elaborates on the reasoning behind proposing DTN as the network
architecture of choice for integrating all different heterogeneous networking
environments composing the Future Internet into a unified networking architecture.
In order to further support this reasoning, an overview of the key DTN aspects is
provided that showcases DTN’s suitability for serving that role. The respective
strengths and weaknesses of this particular integration approach are also discussed.
The next three chapters are dedicated to the presentation of the mechanisms and/or
protocols developed in the context of this thesis. Those chapters are organized
according to the three fundamental Future Internet properties (i.e. Global
Architecture, Global Reach and Global Empowerment) in order to highlight the
association between each developed solution and the respective Future Internet
property; thus, providing the reader with a clearer understanding of the variety and
scope of the activities carried out throughout the development of this thesis. Each of
these chapters provides an overview of the related work that pertains to the
corresponding solution, the necessary theoretical background for understanding the
addressed issues and, finally, a detailed description of the solution’s mechanics and
operational capabilities. In particular, Chapter 3 highlights various data streaming
issues encountered in wireless communication environments and presents Bundle
Streaming Service; a data streaming framework that specifically targets delay
Towards a Future Internet Architecture: Protocols and Support Mechanisms 31
tolerant networking environments, Chapter 4 discusses the strengths and
weaknesses of low-cost, Space-to-ground communication systems and presents a
fully-distributed network architecture that, in contrast to current deployment
approaches, allows for asynchronous communication between end-users across the
globe and, finally, Chapter 5 elaborates on the importance of free Internet access by
disclosing the numerous societal benefits associated with the “Internet for All”
notion introduced by Vint Cerf [11] and presents the Hybrid Packet Scheduling
Scheme; a queueing policy specifically developed for enhancing broadband sharing
networks operational capabilities.
Chapter 6 presents the methodology followed to evaluate both the efficiency and
the effectiveness of the proposed solutions, including the evaluation objectives, the
datasets analysis results, the network scenarios upon which the evaluation process
was based, the metrics used for assessing their performance, and, finally, the tools
used for data analysis and network systems simulation.
Chapter 7 reports the research results of the evaluation activities conducted in the
context of this thesis, which includes the assessment of the proposed solutions’
efficiency and effectiveness under various networking scenarios.
Chapter 8 concludes the thesis with a discussion regarding the overall suitability of
the developed solutions in achieving their goal and provides some future research
directions as well.
32 Chapter 1: Introduction
2. Delay-/DisruptionTolerant Networking as a
Future Global
Internetworking
Architecture
2.1. Current Internet Architecture Limitations
Current deployed Internet protocols have been built based on a number of
fundamental assumptions regarding network topology, network properties (e.g.
packet loss and delay), end-to-end connectivity and users’ behavior. Although valid
at the early stages of Internet’s development, today, most of those assumptions
conflict with some of the operational requirements introduced by emerging
networking environments (see section 1.2). On top of that, the act of employing
those assumptions as design and development guidelines for Internet protocols and
mechanisms have led to imposing several limitations on today’s Internet
architecture capabilities. Due to these limitations, it remains dubious whether the
emerging networking requirements could be sufficiently addressed within the
existing architecture either by means of network “over-dimensioning”, such as
33
34 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
increasing the coverage area of network and communication services, or by
progressively enhancing Internet protocols or mechanisms, through specific and
perhaps conflicting patches.
To illustrate this point, one might consider, for example, how the network topology
assumption regarding the existence of a continuous end-to-end path between
source and destination over the duration of a communication session practically
affects the Internet’s ability to serve emerging networking scenarios. In particular,
several studies suggest that in today’s wireless communications era intermittent
connectivity is an increasingly common phenomenon [12] [13]. This observation is
mainly attributed by the researchers to the increased usage of mobile networking
devices, such as smartphones, tablets, laptops etc. That essentially means that there
are actual use cases where an end-to-end path between the communicating hosts
might not exist. As a result, the TCP/IP stack, which accounts for almost 85% of
today’s Internet traffic [14], is unable to function properly under these
circumstances, since TCP is specifically designed to operate in an end-to-end
fashion; thus, it completely fails when the end-to-end model breaks. Furthermore,
intermittent connectivity might not be solely associated with nodes’ mobility but,
also, with other use-case-specific conditions (e.g. WSN nodes running out of battery
or put into sleeping mode to conserve energy), rendering any over–dimensioning
approaches inappropriate.
The difficulty to cope with intermittent connectivity effectively does not constitute,
of course, the sole limitation of today’s Internet architecture. The limitations, so far,
are further widespread and cover various design and operational aspects of the
current Internet architecture including addressing [15], mobility support [16],
multi-homing, routing scalability, Quality of Service (QoS) provisioning and other
features [17] [18]. Therefore, the Internet architecture, as it currently stands, is
unable to satisfy the requirements of an ubiquitous, mobile and highly
heterogeneous networking environment as the Future Internet is expected to be.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 35
2.2. To “Clean-slate” or not to “Clean-slate”?
Two main approaches have been considered, so far, by the researchers and network
engineers for addressing current Internet architecture limitations.
The first approach, usually dubbed as the “incremental approach”, entails the
evolution of the Internet architecture by implementing incremental patches, while,
at the same time, maintaining backwards compatibility with current Internet
protocols and mechanisms. Several efforts have been put forward, so far, by Internet
Engineering Task Force (IETF) working groups, such as the Network Working
Group (NWG), towards that direction. These efforts have led to the development of
new standards and informational documents that address certain Internet
architecture limitations and/or propose additional network mechanisms for
extending current Internet architecture functionality. Prominent examples of these
efforts constitute the IPv6 protocol [19], which was developed to deal with the
address exhaustion problem faced by IPv4, the Mobile IP [20] and Host Identity
Protocol (HIP) [21] protocols, both of which tackle the problem of mobility and
multi-homing support in IP networks and the E.164 Number to URI Mapping
(ENUM) application [22]: a mapping service between E.164 Numbers and URI
addresses, which can be used for routing purposes between IP and non-IP networks,
such as the Voice over IP (VoIP) and circuit-switched telephone networks.
Although these Internet standards and informational documents manage to address,
at least to some extent, some of the issues pertaining to the Future Internet
architecture, they do not provide a comprehensive solution; thus, they end up
satisfying only a small portion of the Future Internet requirements, as they were
drafted in the motivation section of the present thesis (see section 1.2). Due to this
fact, the Internet community and academia has also considered a more holistic
approach. Contrary to the incremental approach, the holistic approach, usually
dubbed as the “clean-slate approach”, entails a radical redesign of the current
Internet architecture. The solutions developed based on this approach offer new
network service abstractions, through which all the identified issues of the former
state are inherently addressed. Although this might sound ideal, it should be noted
36 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
that the common drawback of clean-slate solutions is the lack of backward
compatibility with pre-existing technologies. Due to this fact, such solutions are
difficult to be accepted by network operators for various reasons including the
massive amount of changes required to the existing infrastructure and the potential
disruption in the operation of already deployed network services and business
models.
Nevertheless, various research projects and initiatives exploring clean-slate Future
Internet architecture solutions have been launched in the last decade [23]. These
projects are not limited to investigating technological issues associated to the
development of the Future Internet architecture but they also investigate social-,
regulation- and policy-related issues surrounding a future internetworked society.
The United States of America (USA) and the European Union (EU) are leading these
efforts so far. In particular, USA has established the Future Internet Design (FIND)
long-term initiative as a part of the NSF NeTS research program, through which at
least six different research efforts are currently funded [24], namely Nebula,
eXpressive Internet Architecture, Named Data Networking (NDN), Mobility First,
ChoiceNet and UltraFlow. The EU, on its part, has also funded and continues to fund
several projects in the context of FP7 and HORIZON2020 programmes, with the
EIFFEL, 4WARD, IOT-i, and EFFECTS+ projects being the most prominent [25].
Furthermore, EU has established the Future Internet Architecture (FIArch) group
[26] as a cross-unit activity that spans across all the EU Future Internet architecture
related projects. The ultimate aims of FIArch are, first, to define a common set of
architectural design principles and, second, to develop an associated reference
architecture of the Future Internet that can guide and unify key technology
developments in the future. In terms of experimental validation and testing, both
American and European research efforts are supported by extended experimental
facilities, Global Environment for Network Innovations (GENI) [27] and Future
Internet Research and Experimentation (FIRE) [28], respectively, which aim to
provide a framework for combining academic research with the wide-scale testing
and experimentation that is required for industry. Apart from the USA and the EU,
Towards a Future Internet Architecture: Protocols and Support Mechanisms 37
other government agencies have also funded relevant research projects and
initiatives with Akari project of Japan, Future Internet Forum (FIF) of Korea and
MOST project of China being the most representative ones [23].
Based on the discussion so far, it is clear that the two approaches have fundamental
differences, which in fact highlight the trade-off between them. Overall, clean-slate
solutions look indeed more promising, but judging from the diverse maturity level
of the aforementioned projects, in conjunction with the amount of research effort
required for them to be completed, it is reasonable to expect that it will take some
time before a universally accepted Future Internet architecture proposal finally
emerges.
Within this context, another evolutionary strategy towards a Future Internet
architecture has been recently introduced in the literature suggesting that
incremental and clean-slate approaches do not have to be competitive, but
complementary [28]. The rationale behind this evolutionary strategy is that
incremental Internet changes should be carefully selected to both enrich current
Internet architecture and address urgent problems, but, at the same time, to serve a
greater plan for transitioning towards a completely redesigned modular network
architecture. As already identified in [17], overlay networks constitute great
examples of this evolutionary strategy, since they represent an elegant way in which
new network functionalities can be implemented based on existing network
technologies to enhance the capabilities of current Internet architecture (e.g. load
balancing enhancement through peer-to-peer networking).
2.3. Delay-/Disruption-Tolerant Networking
Delay-/Disruption-Tolerant Networking is a network architecture that aims to
enable the inter-communication of network units belonging to heterogeneous and
disruptive network environments. It does that by defining an end-to-end, message-
oriented, store-and-forward overlay network, called the “bundle layer”, which exists
usually above the transport layers of the networks it interconnects. Bundle layer can
be deployed in practice by employing the Bundle protocol [7]; hence the name of the
38 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
overlay network. BP constitutes a network protocol that implements all the
necessary functionalities required to support the network operations described in
the DTN architecture IEEE Request For Comment (RFC) document [6].
The operational capability of DTN architecture in transparently integrating
heterogeneous networking environments has already been demonstrated through
various deployments and tested in a wide variety of diverse networking
environments. Examples of such networking environments include interplanetary
[29], military [30], disaster response [31], underwater [32], vehicular [33] [34],
environmental monitoring [35] and some forms of ad-hoc sensor networks [36]
[37]. The growing interest in DTN is further underlined by the continuous
development of delay-tolerant networking protocols and associated standards
under the auspices of the IETF DTN workgroup [38], the Internet Research Task
Force’s (IRTF) DTN research group [39] and the Consultative Committee for Space
Data Systems (CCSDS) DTN workgroup [40].
In this context, DTN is gradually gaining momentum as a promising networking
communication architecture for future global internetworking; a vision that is also
shared and strongly supported by this thesis.
2.3.1. Key Architectural Principles (Pros)
DTN constitutes an evolution of the IPN architecture [41]. As a network
architecture, IPN solely addresses communication needs in Space network
environments, in which long haul delays and frequent communication disruptions
are present [42] [43]. Thus, DTN has been built over a number of principles that are
fundamentally different from those of the current Internet architecture. In essence,
the IRTF DTN Workgroup has further refined the IPN architecture in order to
address the communication needs presented by various challenged and disruptive
terrestrial networks, while still allowing DTN to operate over ordinary networking
environments, such as the Internet. These refined principles though do not render,
by any means, DTN a lesser network architecture; on the contrary, DTN enhances
Towards a Future Internet Architecture: Protocols and Support Mechanisms 39
current Internet architecture capabilities and improves network performance in
various use cases.
RFC document 4838 [6] describes in detail the architectural principles of DTN. The
next eight subsections provide a brief summary of the key principles that make DTN
particularly attractive for supporting a global internetworking architecture.
2.3.1.1.
Store-and-forward model
Store-and-forward model constitutes a forwarding approach that allows DTN to ease
some of the TCP/IP architecture limitations, while still allowing the network to
deliver messages in an end-to-end fashion. With store-and-forward, data are
incrementally moved and stored throughout the network in a hop-by-hop fashion
(i.e. from the previous network node to the next one) until they reach their
destination; thus, the requirement for a continuous end-to-end path between source
and destination over the duration of a communication session ceases to apply. Once
data are successfully moved from the previous node to the next one, they can be
completely removed from the previous node’s memory. In case network
connectivity is not available between the current holder of the data and the next
node, data is stored in memory and is automatically forwarded once the scheduled
or opportunistic communication link gets reactivated. Regarding the way data is
stored in DTNs, in comparison to current Internet architecture, it is envisioned to be
saved in non-volatile memory, enabling applications to cope with restarts after
hardware or other type of failures while network transactions remain pending.
2.3.1.2.
Non-conversational Asynchronous Communication
Due to the assumption of a non-continuous end-to-end path between source and
destination, BP was designed with the philosophy to minimize the number of
communication exchanges required to complete a communication session. In this
context, BP’s data units, called “bundles”, are envisioned to bundle as much as
possible useful information into a single payload data unit (PDU) before it is sent to
the recipient. For example, user credentials along with the actual data payload for a
network service could be merged into a single bundle to avoid both needing two
40 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
round-trips time for completing the communication session and creating two
different PDUs, a tactic that would impose additional communication overhead.
2.3.1.3.
Virtual message switching
In order to accommodate the asynchronous communication principle (see
subsection 2.3.1.2), bundles are structured as variable-length messages with a
theoretical unlimited size, instead of using maximum segment sizes as in TCP.
Having variable-length messages of a theoretical unlimited size serves two main
purposes: First, it creates a message abstraction that is closer to user’s perspective
of communication data making applications development an easier task. Since a
single bundle can incorporate into its data payload anything from a short content
sharing message to a large video file, the application developer do not need to worry
about how she should split the application data units (ADUs) to meet the transport
layer requirements; a rather common problem on network application development
based on the current Internet architecture. Second, this message abstraction enables
the network, as a system, to take more informed decisions regarding routing,
scheduling and path selection by knowing the exact amount of data that need to be
transferred.
2.3.1.4.
Reliability and custody transfer
End-to-end acknowledgements and custody transfer are the two methods provided
by DTN architecture for enhancing delivery reliability in networks where a store-
and-forward forwarding approach is applied.
Regarding the first method, it resembles TCP’s end-to-end reliability philosophy,
and it can be utilized by applications to implement their own end-to-end message
reliability mechanism. The latter method is specifically developed for challenged
network environments in an effort to move retransmission responsibility of bundles
"closer" to their ultimate destination(s); thus, avoiding wasting valuable network
resources, as it would have been the case if every bundle had to be retransmitted
from the source. In particular, custody transfer mechanism allows the delegation of
Towards a Future Internet Architecture: Protocols and Support Mechanisms 41
retransmission responsibility among different nodes along the route path from the
source towards the destination. The node who takes custody of the bundle is called a
custodian. Under various circumstances, a DTN node might refuse to take custody of
a bundle (e.g. due to insufficient storage space).
2.3.1.5.
Routing
The DTN architecture provides a framework for routing and forwarding at the
bundle layer for unicast, anycast, and multicast messages. Therefore, besides acting
as a source, destination or a simple forwarding agent, each DTN node can also
function as a router.
In terms of routing methods, those employed in DTNs differ significantly from their
Internet-based counterparts. In contrast to the Internet-based routing methods that
are able to function properly only over well-connected networks, the routing and
forwarding
mechanisms
communications
patterns,
employed
including
in
DTNs
need
persistent,
to
support
on-demand,
various
scheduled,
opportunistic and predicted communication [6]. This requirement stems from the
large number of diverse networking environments in which DTN may be used as a
communication concept.
The network nodes forming a DTN network might be connected with edges that
follow any of the aforementioned communication patterns. Furthermore, nothing
prevents DTN nodes, in terms of architecture specifications [6], from
interconnecting with other nodes using multiple underlying network technologies.
Therefore, apart from the various communication patterns, those network edges
might also present different and possibly time-varying communication link
characteristics in terms of capacity, delay and communication direction. Due to
these facts, some DTN routing methods are endowed with the ability to defer the
forwarding of a bundle if a better link is known to be available in the future [44]
[45].
42 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
2.3.1.6.
In
order
Naming
to
facilitate
interoperability
between
heterogeneous
network
environments, DTN architecture employs a flexible naming system that is capable of
encapsulating different naming and addressing schemes in the same overall naming
syntax [6].
Forwarding and routing in DTNs is based on DTN endpoints. DTN endpoints are
identified by their associated Endpoint Identifiers (EIDs). Each Endpoint ID is
expressed syntactically as a Uniform Resource Identifier (URI) [46]. Each URI begins
with a scheme name followed by a series of characters constrained by the syntax
defined by the scheme. This latter portion of the URI is called the scheme-specific
part (SSP), and can be quite general expressing different notions, such as Domain
Naming System (DNS) names, expressions of interest or forms of database-like
queries.
Although each DTN node is required to have at least one EID that uniquely identifies
it, multiple DTN nodes might be members of the same DTN Endpoint. Therefore, it is
the responsibility of the scheme designer to define how the SSP of an EID should be
interpreted by a DTN node so as to determine in which node(s) the received bundles
having this particular EID should be forwarded [6].
2.3.1.7.
Late binding
Binding in DTN architecture refers to the interpretation of the SSP of an EID for the
purpose of carrying an associated message towards its destination [6]. With DTN
architecture’s late binding feature, the actual binding process of mapping an EID to a
next-hop EID or to a lower-layer address for transmission can be carried out late in
the delivery process. This is opposed to the current Internet architecture’s early
binding philosophy, where a DNS lookup request should be completed at the source
node to fix the IP address of the destination node prior to the beginning of the
communication session.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 43
In terms of fostering interoperability among heterogeneous network environments,
the late binding feature improves forwarding performance. It does that by allowing
the system to make optimal routing decisions when no end-to-end connectivity
between source and destination nodes is available and helps, as well, reducing the
amount of administrative information propagated throughout the network by
limiting down the mapping synchronization requirements among physical and DTN
addresses to a local topological neighborhood.
2.3.1.8.
Security
The ability of DTN nodes to operate independently as router nodes and store
bundles for long periods of time increases the number of potential attack vectors by
exposing both the overlay network resources and the transferred data itself to
various access-control-related attacks. Although, the employment of endpoint-based
security mechanisms is considered to be sufficient for handling user authentication
and data integrity related issues, it provides no means for protecting the network
infrastructure from unauthorized use. To serve this purpose, DTN architecture
adopts a complementary security architecture, which tries to address the following
issues according to [6]:
•
Prevent unauthorized applications from having their data carried through or
•
stored in the DTN.
•
infrastructure.
•
class of service for which they lack permission.
•
Prevent unauthorized applications from asserting control over the DTN
Prevent otherwise authorized applications from sending bundles at a rate or
Discard bundles that are damaged or improperly modified in transit.
Detect and de-authorize compromised entities.
This complementary security architecture is deployed through the use of Bundle
Security Protocol (BSP) [47]. BSP utilizes both hop-by-hop and end-to-end security
44 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
mechanisms to authenticate DTN nodes as legitimate transceivers of bundles to
each-other and preserve data integrity, as well.
2.3.2. Architectural Shortcomings (Cons)
As previously elaborated, DTN provides a wide set of integrational operating
capabilities and networking features that make it an attractive alternative to
currently-proposed clean-slate Future Internet architectures. Nevertheless, the
research community has identified a number of architectural shortcomings that
puts into question DTN’s viability as an actual internetworking architecture [48].
The most relevant ones to the topic discussed in this thesis are as follows:
(i) Lack of a standardized network and security key management system
The role of network management systems is crucial for guaranteeing the
smooth and secure operation of a network, as network administrators need a
means to monitor the functionality of the nodes, the connections between them,
the services they provide, and preserve, in a scalable manner, network security.
Despite the fact that network security and management as topics have been
successfully addressed in the current Internet architecture, both topics are still
at a primitive stage in DTN. The main reason behind that is the completely
different operational logic of the respective DTN solutions, which is mainly
associated to the long propagation delays and/or intermittent connectivity
characteristics that DTNs present. This different operational logic renders
infeasible the adoption of the corresponding Internet solutions in DTNs; thus,
prolonging the development of the respective DTN solutions. Currently, most of
the proposed solutions regarding the establishment of either a network
monitoring or a security key management system for DTNs are still under
discussion; therefore, there is no yet a standardized network management
system for DTNs.
(ii) BP overhead efficiency
In certain cases, such as data streaming transfers, the BP header imposed on
data flows consisting of small packets might negatively affect the overall
Towards a Future Internet Architecture: Protocols and Support Mechanisms 45
transfer efficiency since the actual data to overhead ratio will be low. BP has a
variable-length header, which can range from a few dozen bytes up to a few
hundred bytes. Apart from multimedia or data streaming transfers, this
particular problem arises in cases where the frame size supported by lower
layer protocols is small, as in the case of IEEE 802.15.4, which is commonly used
on wireless sensor networks.
(iii) Standardization of routing methods
Routing in DTNs constitutes a rather complex issue, given the wide variety of
communication patterns that need to be supported, including persistent, ondemand, scheduled, opportunistic and predicted communication. Although
numerous routing methods have been proposed to date for various
communication scenarios [49] [50], no consensus has been achieved so far
among the researchers working on this topic on which ones of those routing
methods are the most suitable for adopting as standards.
There are quite a few reasons that render decisions regarding this matter a
difficult issue to deal with: First, each method may perform better under
specific conditions. This fact inflates significantly the required standardization
effort by considerably increasing the potential stack of adopted routing
solutions. Second, only a small portion of the proposed routing methods has
been used as the basis for the development of actual routing protocols. Third,
the maturity-level of those methods is quite low. Most of them have been solely
evaluated through simulations, while in a few cases where evaluation included
actual field tests, testing was performed on limited scale networks.
(iv) Lack of adequate Quality of Service support
In contrast to the extensive QoS provisions implemented in the current Internet
architecture, DTN supports three relative priority classes to date, namely the
Bulk, Normal and Expedited classes. A strict priority model is applied between
those three classes (e.g. normal-class bundles are shipped prior to any bulkclass bundles).
46 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
The term “relative” comes from the fact that these priority classes imply
scheduling prioritization only among bundles of the same source meaning that a
high priority bundle from one source may not be delivered faster (or with some
other superior quality of service) than a medium priority bundle from a
different source. Therefore, as it currently defined, the DTN architecture does
not provide any strong QoS guarantees.
(v) Performance of inelastic applications over DTN
Internet applications, such as instant messaging, email or content sharing, are
considered to be elastic network applications, in the sense that they do not have
any real-time delivery constraints, nor do they require an immediate response
from the recipient. That type of applications intuitively works well over DTNs,
in which end-to-end connectivity cannot be taken for granted. On the contrary,
the continuous content storage and retrieval operations performed on each hop
and the lack of any special QoS or forwarding provisions in DTN architecture
might negatively affect the performance of network applications with hard real-
time constraints, such as data or multimedia streaming, even in cases where an
end-to-end path is available [51] [52] [53].
2.4.
Towards a DTN-based Future Internet Architecture
Based on the arguments discussed earlier, one can conclusively highlight three
important points regarding the use of DTN as a primary component for the Future
Internet architecture:
First, it is apparent from the key DTN principles presented in subsection 2.3.1 that
DTN architecture provides all necessary functionality to address in a transparent
manner the majority of current Internet architecture limitations associated with the
support of a ubiquitous, pervasive, mobile and highly heterogeneous networking
environment, such as the Future Internet.
Second, DTN’s flexibility to be deployed as overlay network is in line with the
incremental transitioning logic presented in section 2.2 and allows for further
extension of current Internet architecture capabilities without imposing radical
Towards a Future Internet Architecture: Protocols and Support Mechanisms 47
changes to the underlying architecture; a fact that propels its adoption as a global
internetworking architecture. Therefore, it provides precisely this desirable feature
of addressing urgent problems, while, at the same time, helping towards
transitioning to a redesigned modular network architecture.
The third but crucial point is that despite its rich set of features and ease of
deployment, a small number of currently unresolved issues exist, which prevent
DTN from irresistibly serving the role of the global internetworking architecture.
This fact calls for urgent development of DTN itself.
As a matter of fact, there are several on-going efforts trying to tackle those
unresolved issues. In particular, regarding the shortcomings presented in
subsection 2.3.2, for both the issues of network and security key management, draft
solutions have already been proposed in the form of internet drafts and are
currently at the stage of evaluation before progressing further into a standards track
[54] [55]. In terms of routing, although there is not yet a clear consensus as to which
is the most appropriate routing methods for all the different networking
environments, for some of them, certain solutions seem to stand out. The two most
prominent examples are Contact Graph Routing (CGR) [56], developed in particular
for Space networks, and Probabilistic Routing Protocol for Intermittently Connected
Networks (PRoPHET) [45], which was developed for addressing routing issues in
terrestrial opportunistic communication scenarios. The corresponding protocols for
these two solutions are described in Internet-Draft CGR [57] and RFC6693 [58]
documents, respectively. Supportively, the investigations of other related issues
from the research community, such as the impact of BP overhead on communication
performance in various communication environments [59] [60] and the
performance of inelastic applications over DTNs [8] [61], yield encouraging results.
Overall, previous discussion demonstrates in a clear manner DTN’s ability to
interconnect heterogeneous networking environments into a unified Future Internet
architecture. Therefore, by taking into consideration DTN’s wide range of
operational capabilities, along with the fact that it is currently undergoing a
standardization process [38], it is argued here that DTN framework constitutes a
48 Chapter 2: Delay-/Disruption-Tolerant Networking as a Future Global
Internetworking Architecture
promising network architecture that could serve global internetworking and,
further, support the formation of the Future Internet.
3. Enhancing Data Streaming
Functionality in Delay-/
Disruption-Tolerant
Networks
In order for DTN to be established as a supportive component of the Future Internet,
it should meet Internet users’ expectations for a functionally rich communication
framework. That is, instead of simply providing primary network services, such as
routing, transport and network management, DTN also needs to support various
types of application services beyond simple bulk data transfers. Data streaming
constitutes such an example of application service that has not yet seen much
progress in DTN.
The present thesis deals with this particular type of application service due to its
essential role in today’s Internet and satellite broadcasting world. In fact,
multimedia streaming, a subcategory of data streaming, dominates the Internet
today, as it constitutes the most popular content distribution method. Recent
Internet usage forecasts predict that its dominance will be further established in the
near future, with the IP video traffic accounting for up to 80% of total IP traffic by
2019 [62].
49
50 Chapter 3: Enhancing Data Streaming Functionality in Delay- / DisruptionTolerant Networks
Data streaming over DTNs is a challenging task considering jointly the specific
characteristics of DTN environments and the demanding nature of streaming
applications. In particular, long signal propagation delays, high bit error rates
(BERs), frequent disruptions and variable bandwidth are some of the networking
factors that act inevitably against the basic application principles of data streaming
and call for mechanisms that secure a smoother viewing experience for end-users.
Even in well-connected networks, the impact of store-and-forward practice might
affect streaming performance negatively due to the additional latency imposed by
continuous content storage and retrieval operations performed on each node.
Presently, there are no standard mechanisms available to support this functionality
and typical configurations fail to efficiently transfer data streams. In this context,
BSS is introduced as a practical approach that addresses many of the networking
challenges related to both live and stored data streaming over DTNs.
3.1.
Background
The topic of data streaming in wireless communication environments has been
thoroughly investigated, especially during the last decade. Several research studies
have been published addressing a wide range of data streaming issues for both
terrestrial and Space networking environments.
As far as terrestrial environment is concerned, a substantial number of prior
proposals are targeting MANETs. MANETs are naturally related with terrestrial DTN
in the context of constrained node resources, high-error-rate communication
channels, variable capacity wireless links and mobility dynamics, excluding of
course the end-to-end routing path requirement. In general, the majority of the
efforts focusing on MANETs are moving towards two main directions: efficiency
improvement and redundancy. Among the most popular approaches suggested so
far for improving efficiency are: i) the dynamic optimization of data coding,
throughout the streaming session, so that the encoded bitrate does not exceed the
available bandwidth of the network [63], ii) routing through multiple paths in order
to increase delivery probability [64] [65], iii) packet prioritization to minimize
Towards a Future Internet Architecture: Protocols and Support Mechanisms 51
queuing delay [66] and iv) specifically designated transport layer mechanisms that
aim to reduce delay in the recovery of lost data. Redundancy is usually achieved
through the use of Forward Error Correction (FEC) codes or by applying content
summarization and error-spreading techniques in order to provide error resilience.
A wide range of cross-layer approaches has also been proposed for optimizing video
streaming over MANETs [67].
Unfortunately, due to the fact that the characteristics and deployment objectives of
each type of DTN may vary significantly, most of the aforementioned approaches
cannot be applied efficiently in the context of DTN. In most cases, network functions
such as routing, error recovery and congestion are typically located solely either at
the source or destination, and therefore call for end-to-end connectivity. However,
end-to-end approaches cannot deal with the disruptive nature of DTN; each
individual DTN node should be fully aware of how to handle bundles with frames
that belong to some stream, even when the application per se may be running at the
edges of the network. The use of FEC codes in the bundle layer also has some
drawbacks since it might lead to increased demand for network resources, mainly
bandwidth, without guaranteeing reliable delivery of frames when the coding rate is
not sufficient to replace the losses imposed by the error rate of the communication
channel. Finally, reliability should also be taken into consideration, especially for
critical applications that handle time-sensitive data [68].
The need for data streaming and especially multimedia streaming is also growing
continuously in Space networking environments. This growth comes as a result of
the rapid globalization of the Telecommunications, Media and Entertainment
industry, as well as the increased requirements of the new generation of scientific
Space missions. A number of streaming-specific satellite communication systems
have been proposed using GEO satellites, Medium Earth Orbit (MEO) and LEO
constellations. From a networking perspective, the operation of these systems relies
on streaming-oriented standardized network protocols developed under the
auspices of either the European Telecommunications Standards Institute (ETSI) or
the International Telecommunication Union (ITU). Typical examples of the deployed
52 Chapter 3: Enhancing Data Streaming Functionality in Delay- / DisruptionTolerant Networks
network protocols that provide the satellite-dependent features of the lower levels
of the network stack (layers 1 and 2) are Digital Video Broadcast over Satellite
protocol (DVB-S, versions 1 and 2), the Broadband Satellite Multimedia system and
the Asynchronous Transfer Mode protocol suite. At the same time, common Internet
network protocols are deployed in the upper levels of the stack (layers 3 and above)
[69, 70] in order to provide interoperability at the IP level between terrestrial and
Space applications. In deep Space communications, data streaming is performed
using CCSDS standardized protocols. Packet Telemetry/Packet Telecommand
(TM/TC) or Proximity-1 at the link layer and Space Packet Protocol at the network
layer constitute the most common options [71], while a potentially advanced
streaming functionality should be supported at the application layer. Overall, given
the existing Space protocol stack configurations, it becomes clear that both short
and long range Space communication schemes could benefit from the deployment of
more advanced streaming mechanisms at the upper layers of the network stack.
Yet, the issue of data streaming in the strict context of DTN has not been studied
extensively. In [72], T. Liu and S. Nelakuditi use erasure coding techniques in order
to construct a disruption-tolerant video sequence so that, in the event of disruption,
the network may still provide helpful video content to clients by injecting additional
“summary frames” to the original stream. Another proposal, which is also based on a
coding scheme, is presented in [73] by P. U. Tournoux et al. They introduce Tetrys, a
transport level mechanism based on an on the fly coding scheme which provides full
reliability provided that the encoding ratio used by Tetrys is higher than the average
loss rate. Other well-known approaches employ data replication techniques in order
to improve the availability and the consistency of data despite potentially frequent
network disruptions. One such approach is presented by N. Kumar and J. Kim [74]
who propose a probabilistic data replication strategy, specifically designed for
online video streaming on vehicular DTN. Their scheme aims at increasing the
availability of video streams among a limited group of mobile nodes, whose
connection to the main data server is frequently dropped, by locally replicating the
most common data. Finally, broadcasting systems that rely on delay-tolerant
Towards a Future Internet Architecture: Protocols and Support Mechanisms 53
forwarding of data chunks through mobility of wireless nodes have also been
proposed for distributing multimedia content. In [75], G. Karlsson et al. propose
such a broadcasting system, which follows a receiver-driven approach and provides
unreliable content dissemination to an arbitrarily large group of receivers; under
certain circumstances, mainly related to the mobility of nodes, it can also be used for
multimedia content distribution.
After carefully reviewing the related literature, one can notice that while the
majority of the approaches focus on routing, data replication and coding techniques,
almost none of them employ some type of sophisticated forwarding mechanism;
instead, the role of the existing forwarding mechanisms is typically confined in
simply handling different packet priorities. The approach followed by BSS has its
roots in this observation. In particular, BSS introduces a forwarding technique that
employs a combination of transport services in the forwarding procedure. Similar
ideas already exist in the literature but are employed in a different context: for
example, in [61], N. Panchakarla and J. Ott introduce a hybrid voice streaming
system for terrestrial environments. This system operates in a synchronous manner
based on RTP/UDP transport services, when an end-to-end path exists between the
communicating nodes, and switches to DTN-based voice messaging whenever the
network experiences disruptions/disconnections. In [76], L. Keller et al. propose a
communication system for cooperative video streaming among a group of smart
devices that simultaneously use two network interfaces: cellular 3G/4G to connect
to the video server and Wi-Fi for intra-group communication.
3.2.
Bundle Streaming Service
BSS is a framework that enables ‘streaming’ data to be conveyed via DTN ‘bundles’
in a manner that supports in-order stream processing with minimal latency. Beyond
that, BSS ensures reliable delivery of data to enable ad-hoc ‘playback’ review of
recently received information.
Unlike existing work, BSS emphasizes both the efficient management of available
network resources and the improvement of the end user experience in viewing data
54 Chapter 3: Enhancing Data Streaming Functionality in Delay- / DisruptionTolerant Networks
streams. Given the fact that disruptions are usually localized and experienced only
by a few among many receivers, BSS contributes towards more effective utilization
of available bandwidth by significantly reducing retransmission effort. Another
significant advantage of BSS in comparison to other works is that it does not exclude
the deployment of other sophisticated mechanisms on top of, or below BSS, but
instead, it grafts flexibility that further enhances synergistic application
mechanisms.
Although BSS was initially designed with Interplanetary Internet [41] and its
associated issues in mind, it subsequently became apparent that it could also be
employed over terrestrial DTN environments to improve data streaming
performance, further. In particular, terrestrial environments in which network
nodes either follow a fixed predetermined route or move freely within a dynamic
topology exhibit properties similar to those of Space internetworks, in terms of
bandwidth capacity and connection availability; thus, both environments may be
serviced by a common data streaming solution framework. Potential examples of
real-time applications that could exploit the capabilities provided by this framework
are one-way voice, video or continuous telemetry streaming.
Presently, BSS is implemented as part of the Interplanetary Overlay Network
platform (ION) [77], a well-known software distribution by NASA that implements
the DTN architecture as described in RFC 4838 [6], however, there are also plans for
porting it to other widely-used DTN platforms including DTN2 [78] and IBR-DTN
[79] in the near future.
3.2.1. Design and Implementation Details
The design of BSS was based on the streaming-specific network and application
functionalities envisioned for a DTN-based streaming service. In particular, the
functionalities of in-order delivery of bundles and reliability were considered to be
of paramount importance: In-order delivery is considered a crucial property of data
streaming applications, since it is responsible for the viewing experience of the end-
user. Unfortunately, DTN messaging protocols, such as BP, do not include any
Towards a Future Internet Architecture: Protocols and Support Mechanisms 55
inherent mechanism for in-order delivery of bundles. Therefore, this issue had to be
specifically addressed by BSS’s design. Reliability also constituted an issue of
concern given the potential use of BSS in critical terrestrial and Space applications
that are designed to handle time-sensitive data.
Considering jointly BSS framework’s requirements and implementation details,
another fundamental issue that needed to be addressed during the design phase
was the stack layer at which BSS functionality should be implemented. One might
argue that implementing BSS solely at the application layer constitutes a flexible
solution that allows for totally shifting control over the entire range of BSS features
to the application user. Although such an argument might seem reasonable, it also
presents a number of fundamental drawbacks. The BP specification [7] makes it
clear that the application layer is unaware of the supported transport layer services
underlying bundle transmission; the bundle layer is responsible for selecting the
transport layer service, through which the bundle will be forwarded to the next
node. Furthermore, BP specification does not address the issue of how to select the
appropriate transport layer service. Currently, all major DTN stack implementations
follow a one-to-one approach that binds transmissions to specific endpoints to
predetermined underlying protocols. Implementing BSS solely at the application
layer, in-line with the aforementioned one-to-one approach, implies, inevitably that
the BSS application framework is associated with specific transport services.
Guaranteeing reliability in such cases can either be achieved by an underlying
reliable transport service or by application enhancements that supplement a best-
effort transport service; in the latter case, retransmission of lost bundles becomes
clearly an application task. The two options entail a trade-off. The additional
overhead that characterizes a fine-grained reliable transport protocol has the effect
of reducing the volume of data that must be retransmitted, while the lack of a
reliable transport service may result in retransmission of entire bundles by the
application. Moreover, both options entail the additional disadvantage of either
retarding data delivery until retransmission is completed or else foregoing in-order
delivery of data to the user.
56 Chapter 3: Enhancing Data Streaming Functionality in Delay- / DisruptionTolerant Networks
In order to compensate for this discrepancy and guarantee reliability without
increasing delivery latency, BSS balances the aforementioned trade-off, within the
“bundle” layer, by delivering streaming data in transmission (rather than reception)
order. This is achieved by employing a pair of transport services; a best-effort
service together with a reliable service. The best-effort service delivers most of the
data in transmission order with minimal latency and minimal initial transmission
overhead, while the fine-grained reliable transport service ensures eventual
delivery of all data lost in transit while minimizing retransmission overhead.
Moreover, because the BSS design is neutral with respect to quality of service (QoS)
markings defined in the BP specification [7], it naturally supports concurrent
transmission of multiple streams at different priorities over the same
communication link.
All network functionality of BSS is placed at the bottom of bundle layer in the form
of a convergence layer protocol. Application-related functionalities, such as reordering of “media” packets received out of transmission order, are preserved at the
application layer and are implemented in the form of a software library. The
architecture of BSS is depicted in Figure. 3-1.
Figure 3-1. Bundle Streaming Service architecture.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 57
3.2.2. Forwarding Algorithm
BSS specifically targets DTN, thus its forwarding decisions are made on a hop-by-
hop basis. The BSS forwarding algorithm is executed by the BSS convergence layer
protocol (BSSP) daemon each time a new bundle is received. Bundle creation time is
the criterion on which forwarding decisions are based. The BSSP daemon keeps
track of the creation times of the bundles flowing from a node to the next one, at
every “hop” along the end-to-end path. Each of the forwarder node's neighbors must
support two inducts, one for a "best-effort" protocol, such as UDP and another for a
reliable protocol, such as TCP. BSSP daemon selects the appropriate outduct for
forwarding the bundle pending dispatch by applying the following rule:
“Each bundle whose creation time is greater than that of all other bundles seen on this
stream is forwarded via the ’best-efforts’ outduct. All others are forwarded via the
‘reliable’ outduct.”
BSSP implements its own reliability mechanism; thus, every bundle sent by BSSP
has to be acknowledged. If the acknowledgment for a bundle does not arrive prior to
timeout, the bundle is re-transmitted up to X times — a parameter set either by the
application or by the network administrator. Since the creation time of a retransmitted bundle cannot be greater than that of all other bundles previously
transmitted in this particular stream, it is instead forwarded to the reliable outduct
for transmission. In general, BSSP needs to run in every intermediate node of the
end-to-end path. However, the potential absence of BSSP does not pose any serious
problem. The BSSP-unaware node simply does not identify the flow of bundles as
BSS traffic and treats them as normal traffic; at worst, it may impose some
additional latency. That is, forwarding traffic based on BSS logic has the potential to
enrich streaming services but its partial deployment does not really cause any
damage.
Overall, the design of BSSP ensures that all bundles in the stream are eventually
delivered to their final destination, and all bundles that were never lost or
retransmitted along the path are delivered in transmission order with minimum
58 Chapter 3: Enhancing Data Streaming Functionality in Delay- / DisruptionTolerant Networks
latency. The flow of streaming data never waits for a retransmission to succeed
since such an action can degrade the viewing experience of a participant in a data
streaming session. In the event that a bundle sent over the non-reliable transport
service does not arrive at its next-hop node, that bundle simply will not be included
in the flow of streaming data displayed at the destination. Gaps might appear in the
display but never any regressions.
Once a lost bundle gets re-forwarded, successfully, it is identified as out-of-stream
since its creation time is earlier than that of subsequently forwarded bundles. It
eventually ends up at the destination, but because it is out-of-stream, it does not get
included in the displayed flow of streaming data; it just goes into the database, along
with all of the successfully streamed data that has already been displayed. This
enables the user to employ the playback features of the streaming service library to
rewind back through the database and replay the data that were missed in sequence
with the data that originally were successfully received.
3.2.3. Display Algorithm
The receiver’s application is built using the BSS library. The intention is to enable
applications to pass streaming data received in transmission time order to an
application-specific "display" function but also to store all received data (including
out-of-order data) in a private database for playback under user control. The
reception and real-time display of in-order data is performed by a background
thread, leaving the application's main thread free to respond to user commands
controlling playback or other application-specific functions.
Whenever the background thread receives a bundle, it inserts it into the BSS
database in creation-time order for replay on demand. It also checks bundle
creation time in order to decide whether to pass the bundle to a callback function
for real-time display or to some other function for further stream processing.
Meanwhile, the main thread of the application may respond to user commands by
calling BSS library functions. For example, library functions may retrieve data from
the time-ordered database and allow operations such as forward, backward, fast-
Towards a Future Internet Architecture: Protocols and Support Mechanisms 59
forward, freeze etc. The result of the above-described process is unimpeded real-
time streaming of all data delivered in transmission order, together with
comprehensive replay and review of all the data in the stream.
60 Chapter 3: Enhancing Data Streaming Functionality in Delay- / DisruptionTolerant Networks
4. Extending Information
Access through an
Asynchronous, Spaceoriented Data
Dissemination System
Recent trends in aerospace engineering include the development of smaller, lowcost satellites, enabling small-scale organizations with limited budget, such as non-
profitable initiatives, universities, research institutes, startups, as well as small and
medium-sized enterprises, to build their own spacecraft [80]. Likewise,
corresponding ground communication equipment is inexpensive and easy to set up,
allowing for extensive deployment that leads to more frequent contact
opportunities, albeit of considerably lower capacity. These new technologies are
bound to extend information access beyond any geographical boundaries — or, as
dictated by the Global Reach property, “wherever information may be needed or
produced” — by fostering the deployment of cost-effective satellite networking
systems.
Nevertheless, in order for this vision to be realized, a networking framework is
61
62 Chapter 4: Extending Information Access through an Asynchronous, Spaceoriented Data Dissemination System
required that allows Space assets to be seamlessly incorporated into the Future
Internet architecture. In this context, a fully distributed, LEO-based data
dissemination system is proposed as the means to implement such a networking
framework and extend traditional satellite communications to enable asynchronous
delivery of data to interested end-users across the globe [9].
4.1.
Background and Design Guidelines
Deploying satellites in LEO presents several advantages including high proximity to
the Earth’s surface, low transmission power requirements and lower launch costs in
comparison to GEO satellite deployments. From a networking perspective, the main
benefits of LEO satellite communications are closely associated to the low-path
delay that they offer, typically about 20-25ms. This fact contributes to high service
quality links ensuring seamless integration into terrestrial networks, while, at the
same time, allows for compatibility with TCP/IP-like protocols used for terrestrial
applications; thus, supporting full adaptation to interactive applications with real-
time constraints. Despite those advantages, LEO satellite communications are
susceptible to frequent communication disruptions, a liability originating from their
small coverage area footprint. In order to address this discrepancy, the deployment
of constellations of satellites with inter-satellite links is usually considered for
maximizing connectivity opportunities, preserving already established end-to-end
connections between terminals and mitigating the problem of long revisit periods
inflicted by LEO orbits.
Given the high complexity and costs associated with implementing inter-satellite
links, the employment of disruption-tolerant networking technologies could also
pose an interesting and cost-effective alternative solution, due to their ability to
cope with intermittent communication channels and lack of end-to-end connectivity.
The advantages of applying the DTN architecture [6] to LEO satellite
communications have already been showcased in previous works. In [81], the
authors have shown, by emulating various LEO satellite communication scenarios,
that certain DTN technology features could significantly boost LEO satellites
Towards a Future Internet Architecture: Protocols and Support Mechanisms 63
communication performance in comparison to conventional communication
approaches. In particular, they showed that proactive fragmentation and store-and-
forward features could counteract link disruptions frequently met when using
mobile terminals and remedy possible issues with respect to lack of free channels
during handovers between satellites. Besides emulation studies though, DTN
communications have also been used in actual LEO satellites [82][83]. The results
obtained in the context of the United Kingdom – Disaster Monitoring Constellation
(DMC) satellite DTN experiments clearly showcased the performance advantages of
proactive fragmentation, especially when performing large file transfers over
multiple ground stations.
Various LEO satellite constellations have been proposed and deployed to date, both
for commercial and scientific purposes. Iridium, Globalstar and OrbComm systems
are the main representatives of the commercial satellite telecommunication market
that base their operations on LEO satellite constellations. Their aim is to provide
voice and data communication services capable of supporting a diverse range of
applications including geodesy, navigation, remote sensing, personal and asset
tracking, data monitoring and broadband networking applications [4]. The number
of satellites currently deployed in each of these constellations varies from 5 to 66,
with Iridium and Orbcomm systems providing global coverage. Each of these
systems provides its own benefits, with Iridium presenting the best performance
and providing the lowest latency in delivering messages, whereas the Orbcomm
system is best suited for small amounts of data communications. Other commercial
satellite systems, such as the RapidEye and DMC, also follow a similar approach. In
particular, RapidEye and DMC systems rely on a constellation of 5 and 6 LEO
satellites for their operation, respectively [84][85]. These systems are mainly
offering imaging services for both humanitarian and commercial applications
focusing on providing customized solutions for agriculture, forestry, infrastructure,
security, and emergency monitoring applications. Furthermore, a new round of
initiatives have been launched recently by commercial ventures, such as SpaceX,
Qualcomm, Virgin group, AirBus etc., on developing even more extended LEO
64 Chapter 4: Extending Information Access through an Asynchronous, Spaceoriented Data Dissemination System
satellite communication systems in comparison to the ones described above,
composed of hundreds or even thousands of micro satellites [86].
At the other end of the applications spectrum, various scientific Space missions have
also adopted a constellation-based operational approach in order to increase their
coverage area, reduce communication delay and enable near-coincident
observations. The Morning and Afternoon-Train (A-Train) constellations, mainly
comprising of satellites belonging to NASA's Earth Observation System, as well as
ESA's Sentinels missions constitute the most prominent examples. Their main aim is
to provide Earth observation services, including global monitoring of the land
surface, biosphere, atmosphere, and oceans, as well as emergency response and
security services. The number of satellites on each constellation varies from 2 to 6
satellites, while their altitude varies from 693km to 824km [87][88][89].
Regarding their network architecture, a typical characteristic of the currently
deployed LEO satellite constellation systems is the application of a monolithic
deployment philosophy, according to which a few high-end satellites are placed at
different points around the globe with a single satellite occupying each orbital
location. This particular approach carries several drawbacks including: i) relatively
high launching costs, due to the great size and heavy weight of the satellites, ii) long
deployment periods, since monolithic constellations require most of the units to be
in orbit before being capable of providing an acceptable level of service, and iii)
single point of failures, since the disruption of service of a single satellite can cause
the total disruption of the constellation service. Another typical characteristic of the
currently deployed LEO satellite constellation systems is the employment of pointto-point links as their primary transmission medium for their downlink services.
These point-to-point links are usually supported by data relay systems, such as the
Tracking and Data Relay Satellite System (TDRS), and are commonly based on Ku
and S bands [90][91]. Although these point-to-point links present superior
performance, in the order of hundreds of Mbps, they usually entail high
administration overhead, since all contacts with ground stations and relay satellites
need to be prescheduled. At the same time, the deployment of point-to-point links
Towards a Future Internet Architecture: Protocols and Support Mechanisms 65
considerably increases the development cost of ground stations, since more
sophisticated equipment is required to support the high data transmission speeds
offered by those links. Broadcast downlink services are provided as well, albeit
rarely, either as backup downlink services or for real-time, locally confined
consumption of Space data [92]. Known examples of satellites that provide
broadcast downlink services include NASA’S EOS Terra, Aqua and NPP missions,
which employ the Direct Broadcast technology developed by NASA's Direct Readout
Laboratory.
In order to overcome the majority of these problems, a recent trend has emerged,
which calls for the adoption of a distributed deployment and data dissemination
philosophy. In this new approach, each of the high-end satellites is replaced by a
swarm of small, inexpensive satellites that employ broadcasting as their primary
method for data transmission. The adoption of such an approach allows for the
launching and development costs to be significantly reduced, due to the lighter
weight of the satellites and the use of off-the-shelf components. Additionally, it
enhances performance during the deployment phase, since smaller satellites can be
launched more quickly and can be arranged to provide a high level of service before
the entire cluster is in place. At the same time, the robustness of the system is
largely improved, since satellite clusters can continue to provide service even when
a few units have failed.
As an extension of this innovative deployment and data dissemination practice in
Space, the additional deployment of numerous low-cost ground stations has also
been proposed as a means to enhance timely reception of data by increasing the
number of downlink opportunities. On top of that, the adoption of a peer-to-peer-
based or multicast distribution scheme could further enhance the on-ground data
distribution process by enabling fragmented downlinked data received from
different ground stations to be universally distributed in an efficient manner across
every interested party. A similar approach is adopted in [93], where the authors
propose a satellite multicast overlay system for existing terrestrial content
distribution networks to improve performance in video applications. Their proposal
66 Chapter 4: Extending Information Access through an Asynchronous, Spaceoriented Data Dissemination System
is based on the fact that satellite networking has distinct advantages over terrestrial
networks in being able to serve multiple, delay-insensitive requests simultaneously
by distributing high bandwidth content across wide geographic areas. Due to this
unique characteristic, satellite for multicasting is used to leverage the scalability of
the proposed system with increasing demand.
In practical terms, a deployment philosophy that closely resembles this distributed
approach will be applied in a massive scale for the first time on the QB50 mission; a
future Space mission, originated by the corresponding EU FP7 research project [94].
QB50 aims to be the first Space mission to deploy a cluster of 50 Cubesats for multipoint, in-situ measurements in the lower thermosphere. These 50 Cubesats will be
sequentially deployed in a nearly circular orbit at 320 km altitude and the Global
Educational Network for Satellite Operations (GENSO), a network of low-cost
ground stations spread across the globe, will be employed to provide the necessary
downlink services.
4.2.
A Fully Distributed and Asynchronous Data Dissemination
Model
Departing from these design guidelines, the proposed data dissemination model is
depicted in Figure 4-1.
Figure 4-1. Data return model with multiple end-users receiving data through
broadcasting.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 67
According to this model, data transmission from Space is performed through a best-
effort, broadcast transmission approach with optional delivery feedback. Data is
received on ground by the end-users. As it is depicted in Figure 4-1, not all endusers are required to operate a ground station (GS). Furthermore, few dedicated
ground stations are assumed to have bidirectional links with the satellites in order
for the optional delivery mechanism to be supported and the upload of new content
to the satellites to be enabled as well.
A number of Space communication protocols standardized by CCSDS may facilitate
the transmission of data from Space. Common data-link protocols include the
TM/TC [95][96] and Advanced Orbiting Systems (AOS) [97]. On top of these
protocols files can be transferred using the CCSDS File Delivery Protocol (CFDP)
[98], DTN BP or even conventional terrestrial protocols, such as TCP/IP.
Once the data reach the ground, they can be immediately exploited by the end-user
that has received it. From this point on, the particular end-user can advertise the
received data to other interested parties through the subscription service. Further
dissemination of data on the ground is performed through a peer-to-peer multicast
distribution approach, similar to a push variant of the popular BitTorrent protocol
[99]. Adopting a peer-to-peer multicast approach targets at eliminating duplicate
network traffic and offloading central nodes, traditionally burdened with the role of
collecting and distributing Space data. Moreover, acquired data can be readily
available, through such a distribution scheme, to interested end-users located in the
vicinity of the receiving ground station, avoiding any delays caused by central
collection and processing.
Considering the available network architecture options, DTN architecture seems to
be a more natural fit to support this dissemination model, as it can facilitate
transmission of satellite data from Space all the way down to the interested end-
users and backwards. DTN inherently addresses a number of issues that arise in
such a networking scenario, while it is general enough to allow for appropriate
extensions when necessary.
68 Chapter 4: Extending Information Access through an Asynchronous, Spaceoriented Data Dissemination System
The following section crystallizes the deployment details of the proposed data
dissemination model and presents a short analysis on how the design choices
described above can be addressed with DTN.
4.3.
A DTN-based Space-to-ground Data Dissemination System
In order to allow end-users to subscribe to certain types of service one DTN EID is
created for each type of service. From an end-user perspective, these types of
services might represent different types of data (e.g. educational content, news,
alerts etc.).
Additionally, each end-user has its own unique EID, allowing for personalized data
delivery. ADUs are created on a per-type-of-service basis and are then fragmented
into multiple bundles according to the duration of the opportunity windows. As
depicted in Figure 4-2, each ADU and, consequently, the bundle fragments belonging
to it are sent to the predefined EIDs, depending on the type of service or unique EID
of the included payload.
Figure 4-2. DTN deployment for implementing a distributed data return model.
End-users are submitting requests for joining and leaving the EIDs of their choice to
a subscription service on the ground network. The subscription service exposes the
group EIDs to the individual EIDs contained in each group and compute data routes
Towards a Future Internet Architecture: Protocols and Support Mechanisms 69
based on certain network metrics received by the end-users. End-users contact the
subscription service at regular intervals and setup routes for forwarding incoming
bundles depending on their EID. The subscription service may advertise alternative
routes for enhanced reliability.
Reliability for the Space segment, in case the system supports uplink data transfers,
is added using the custody transfer mechanism provided by the BP. Since there is
usually a single ground station responsible for sending data to a certain satellite, all
custody acceptance reports will need to be routed through the delegated ground
station. Therefore, when a ground station receives a bundle, it will produce a
custody acceptance report that will be routed to the delegated ground station and
remain there until the next uplink opportunity. When the delegated ground station
itself receives a bundle, the created custody acceptance report will be locally stored
until the next uplink opportunity. A contact based routing algorithm is the suggested
solution for the uplink communication. Reliability for the ground segment will be
based on TCP and, therefore, no special provisions are necessary.
4.3.1. Data Fragmentation
The BP receives ADUs from the application layer and transforms them into DTN
PDUs called bundles. The main objective of the proposed architecture is
transporting these ADUs to the interested end-users. The abstract protocol stack at
the satellite, ground station and end-user is depicted in Figure 4-3.
Application
Application
Bundle Layer
Space protocols
Bundle Layer
(application)
Space
protocols
Internet
protocols
Bundle Layer
Internet protocols
Figure 4-3. Abstract DTN-based protocol stack for each network entity.
70 Chapter 4: Extending Information Access through an Asynchronous, Spaceoriented Data Dissemination System
Each ADU received by BP layer at the satellite may be encapsulated in one or more
bundles, depending on the ADU size and the nature of the contact opportunities.
Even though DTN has provisions for fragmenting a bundle in a number of cases
during the transfer, this may not be necessary, as the original bundle size will be
small enough to be transmitted over the Space link. Once on the ground network,
the size of the bundle does not affect bundle delivery.
4.3.2. Addressing
The flexibility of the addressing scheme defined in the DTN architecture allows for
easily accommodating the data communication needs of the presented data
dissemination model. Specifically, each ADU on board the satellite will be destined
to a predefined endpoint containing the nodes interested in this type of data. Endusers will register for the relevant endpoints according to their individual needs.
4.3.3. Routing
Data transmission from the ground station to the satellite, which is based on point-
to-point links, can be naturally accommodated by DTN’s off-the-shelf CGR.
Regarding the opposite direction, where the satellite broadcasts data over a large
area and the actual receivers may be unknown, no special routing provisions need
to be made.
Once on the ground, data are delivered to the interested end-users over a terrestrial
distribution scheme that follows the same principles as the BitTorrent architecture
[99]. Concepts from IP multicasting [100] or transport layer multicast approaches
[101] can be borrowed to form the bundle-layer multicast network. In this case,
since routers at the IP layer have no knowledge of the types of service available at
the bundle layer, a separate protocol at the application layer might be necessary to
facilitate the setup of the distribution trees.
4.3.4. Reliability
Employing the BP for the LEO-based data dissemination model studied in this thesis
would require adding reliability only for the Space link between the satellite and the
Towards a Future Internet Architecture: Protocols and Support Mechanisms 71
ground stations. Once the data reach earth, traditional Internet protocols (i.e. TCP)
will provide the necessary facility for guaranteed bundle delivery. In the case where
uplink opportunities are available, it would be very beneficial for the satellite to
have information on successful bundle delivery at some ground station, so that it
will refrain from retransmitting already received bundles. This can be accomplished
by either accepting custody at the ground station or by utilizing the “Report When
Bundle Received” bundle delivery option. In either case, the satellite will be able to
safely delete the bundle for which custody transfer or bundle received report has
been received, both freeing valuable storage space, as well as avoiding redundant
bundle retransmissions. In case the custody transfer mechanism is used, it is
possible for multiple ground stations to have custody for the same bundle due to the
broadcast nature of the Space link communication. Even though this is a rare case,
the DTN architecture supports joint custody by multiple nodes. Finally, due to the
inherent support of DTN for asymmetrical communication links, the custody
transfer or bundle received reports can be routed to the satellite via a ground
station other than the one that has actually received the bundle, supporting the
single uplink ground station limitation present in many occasions.
72 Chapter 4: Extending Information Access through an Asynchronous, Spaceoriented Data Dissemination System
5. Widening Internet
Accessibility through
Broadband Sharing
As presented thoroughly in section 1.2., widening Internet accessibility can be
associated with multiple societal benefits, such as reducing the negative
consequences of digital exclusion faced by less-privileged people [102], fighting
privacy infringement practices, supporting emergency services, providing
ubiquitous networking and benefiting economic growth. In-line with the Global
Empowerment property and “The Internet is for Everyone” notion introduced by
Vint Cerf [11], free Internet access could be used as a powerful substrate towards
widening Internet accessibility by extending information access to a much wider
audience.
Recently, new proposals and initiatives have been put forward to extend the original
scope of broadband connection sharing in order to be used as a framework for
providing free Internet access [103] [104]. The fundamental idea behind these
proposals and initiatives is to consider broadband sharing as a resource pooling
service. Through this service, the unused capacity of a broadband connection could
be leveraged to carry the traffic generated by guest users. The deployment of the
73
74 Chapter 5: Widening Internet Accessibility through Broadband Sharing
broadband sharing scheme itself could be based on the well-known concept of
UPNs, whereas the free and open-access Internet service could be implemented by
mandating lower-priority access to available resources. A practical example of such
open-access broadband connection sharing scheme was demonstrated in the
context of Public Access WI-FI (PAWS) research project [105].
Following the principles of UPNs, the concept of open-access domestic broadband
connection sharing comprises two main entities: i) home users, who act as microproviders for guest users by building an access network and sharing their Internet
connection and ii) guest users, who wish to freely access the Internet by exploiting
the available unused capacity.
The network dynamics developed between these two entities are rather complex
considering the diverse set of requirements that need to be met. In particular, from
home users’ perspective, domestic broadband connection sharing schemes should
preserve sufficient quality of network experience, similar to the one they enjoyed
prior to sharing their connection. Unlike home users, guest users may not expect such
high-quality service; rather, in-line with the philosophy that governs domestic
broadband open-access services, a lower level of service should be satisfactory.
However, this level cannot be so low as to violate the notion of service itself. Besides
the complexity of technically accommodating guest-user traffic at minor cost to
home users’ network performance, the dynamics of guest users need also be
regulated: for example, scalability, which is defined as the ability to serve more than
one guest user at a time, becomes a critical concern. Therefore, a successful
domestic broadband connection sharing scheme should be able to guarantee high
quality of service for home users and accommodate as much guest users as possible.
Both the quality of network experience enjoyed by home users and the amount of
resources allocated to guest users are directly affected by the respective scheduling
and dropping policies applied in the broadband connection sharing service. The
DiffServ framework is commonly used to enforce such policies: each IP packet is
marked, according to its priority, by appropriately manipulating IP header’s 6-bit
differentiated
services
code
point
(DSCP).
Since
the
entire
DSCP
Towards a Future Internet Architecture: Protocols and Support Mechanisms 75
marking/processing is currently done in the Broadband Remote Access Server
(BRAS), it becomes apparent that some sophisticated traffic control should also be
performed at the edges of the network.
In this context, a numerical analysis is conducted to investigate the impact of guest-
user traffic on home-user traffic at the edges of the network in terms of queueing
delay and average system time, in association with different packet-size
distributions and varying network speeds. Then, based on the findings of the
analysis, Hybrid Packet Scheduling Scheme is introduced, as a queue-management
framework that allows for a more effective exploitation of the available network
resources by guest users.
5.1.
Background
Several commercial and non-commercial initiatives have already explored sharing
users broadband Internet connection via various wireless technologies. FON [106],
Open Garden [107], OpenSpark [108] and Open-Mesh [109] are among the most
popular broadband sharing schemes currently operating across the globe. Although
those schemes are gaining worldwide acceptance, their design and deployment
philosophy of extending users’ paid services confines their operation as free and
open-access broadband connection sharing platforms.
Unlike the aforementioned initiatives, Open Wireless Movement [104] and PAWS
[105] introduce a new Internet access paradigm based on a set of techniques that
enable guest users to exploit the spare capacity in domestic broadband networks by
allowing Less-than-Best-Effort (LBE) access to them. In particular, both schemes
adopt an approach of community-wide participation, where broadband customers
are enabled to donate a portion of their high-speed broadband Internet connection
for use by fellow citizens.
LBE service, also known as scavenger class of traffic, was first introduced almost two
decades ago, as an additional class of Best-Effort (BE) service supported by the
Internet2 network. Its original aim was to provide a lower access priority service
compared to the standard Internet BE service for loss-, delay- and jitter-tolerant
76 Chapter 5: Widening Internet Accessibility through Broadband Sharing
applications, such as bulk data transfers and network backups. Later, it was
envisaged to support non-commercial traffic transfers, as well as other network
services and applications, such as network signaling and gaming.
Several packet-scheduling approaches for the deployment of LBE service on
Internet Service Providers’ (ISPs) core network have been considered so far. In
[110], Carlberg et al. present a degraded service model for IP networks that
supports LBE traffic and introduce the Persistent Class Based Queueing system (P-
CBQ). P-CBQ degrades the service of a class at some specified rate by dropping its
packets, according to a penalizing algorithm. Following the incorporation of the
scavenger class of traffic in the DiffServ architecture, researchers working on the
development of well-known educational networks, such as GEANT and Internet2,
studied various class-based packet-scheduling policies as possible candidates for
supporting it. They concluded that, in most cases, assigning a small fixed weight to
the service of the LBE traffic constitutes the most appropriate configuration
[111][112].
Regarding the deployment practices for LBE service at the edges of ISPs’ network
(e.g. access points of domestic broadband connections or user devices), the
proposed approaches can be classified into two generic categories. The first
category includes all approaches that are based on non-intrusive transport
protocols [113], such as the Low Extra Delay Background Transport protocol
(LEDBAT) [114]. These approaches are employed on a per-application basis and are
mainly used to transfer certain application data as background traffic without
hampering the network performance of other delay-sensitive Internet applications
run by the user. Apart from their typical role though, their use in resource sharing
scenarios has also been considered. In particular, the authors of [115] explore the
capabilities of LEDBAT to carry non-commercial traffic by exploiting unused 4G
satellite link capacity. The second category includes approaches that are based on
specialized packet scheduling policies, including different priority queueing and
weighted fair queueing variations. These are employed on a per-gateway basis and
their main purpose is to provide guest users with a lower-priority access to the
Towards a Future Internet Architecture: Protocols and Support Mechanisms 77
network resources in contrast to priority level offered to typical subscribers. One of
the most sophisticated methods specifically developed for enabling guest users to
freely access the Internet, while imposing minimal impact on sharers’ network
performance, is UPN Queuing. In [116], Psaras et al. introduced UPN Queuing
(UPNQ), a packet-scheduling algorithm based on the non-preemptive PQ scheme. In
particular, UPNQ regulates the service rate of guest-user traffic by taking into
account its percentage impact (k) on home users’ average queuing delay. The
authors show, both experimentally and analytically, that by employing UPNQ, a
small amount of LBE traffic (i.e. guest-user packets) can be served with statistically
zero impact on the network performance enjoyed by home users. As initially shown
in [117] though, this percentage-based approach entails a significant drawback; it
does not capture any quantitative characteristics of the additionally imposed
queueing delay. For example, although, under some certain traffic load, the
percentage impact on home users’ queueing delay might be identical in two different
access networks, the respective temporal impact may differ by several milliseconds.
Although methods of both categories can be used supposedly for sharing-resource
purposes, in practice, forcing guest users to use either non-intrusive transport
protocols or specific-purpose applications is technically more challenging and might
as well discourage them from actually using the service. What would be more
practical is to transparently control system dynamics between home- and guest-
user traffic by employing appropriate queue-management policies at gateways,
placed at network edges. That said, the present thesis focuses specifically on this
topic.
5.2.
Broadband Sharing System Modeling
In a UPN, traffic can be classified into two major categories: home- and guest-traffic.
For both groups the following are assumed: i) traffic is generated by a large number
of flows, ii) flow arrivals follow a Poisson distribution, iii) packet-size varies from
40-1500 bytes and iv) same packet-size distribution.
78 Chapter 5: Widening Internet Accessibility through Broadband Sharing
Traffic differentiation in an access point that implements a network resource
sharing scheme can be aligned to a queue-management perspective by assigning
each group of traffic to a different class of service. The following traffic-classes are
considered: i) BE-class and ii) LBE-class. Packets of the home-traffic are assigned to
the BE-class, while guest-traffic packets are assigned to the LBE-class.
Each micro-provider is expected to share a single broadband link, so it is reasonable
to assume that the investigated system is supported by a single server. To simplify
the analysis, buffering space of the access point is assumed to be infinite. In this
context, the ΣΜ/G/1 modeling system is selected as the basis for the analysis.
5.2.1. Numerical Analysis
Through the present analysis, the aim is to evaluate the impact of guest-user traffic
on home-user traffic in terms of queueing delay and average system time.
Bandwidth availability on the transmission link and packet-size distribution of the
flows are two critical factors for the overall behavior of the system, since their
respective values have a direct impact on the average service time. In general,
service time is proportional to the size of a packet. Based on the assumptions
presented in the previous section, it is expected from both classes to present equal
average service times.
In particular, the primary interest of the analysis is in drawing conclusions for the
upstream link, since its allocated bandwidth is usually low and highly variable.
Typical values of the allocated bandwidth for the upstream link range from hundred
Kbps to a few Mbps. Unlike upstream, the downstream link is significantly faster,
with typical bandwidth values in the order of tens of Mbps.
As far as packet-size distribution is concerned, the particular choice of queueing
model allows for investigating the behavior of queueing delay under general
distribution service times. That said, the impact of different packet-size
distributions on the queueing delay of the system is analyzed.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 79
Table 5-1. List of the Examined Packet-size Distributions.
Model Descriptions (in bytes)
Distribution
40-576-1500 (tri-m)
58% - 33% - 9% [118]
40-576-1500 (tri-m)
40% - 20% - 40% [119]
64-1500 (bi-m)
60% - 40% [118]
64-1300-1500 (tri-m)
Symbol
λBE
λLBE
λ = λΒΕ+λLΒΕ
60% - 20% - 20% [119]
Table 5-2. Notation Table.
Description
Arrival rate of Best Effort class (home-user traffic)
Arrival rate of Less than Best Effort class (guestuser traffic)
Total arrival rate
E(SBE), E(SLBE)
Average service time of each class
ρBE = λBE* E(SBE)
Utilization of Best Effort class
E(S)
Average system service time
ρLBE = λLBE* E(SLBE)
Utilization of Less than Best Effort class
Cv(SBE), Cv(SLBE)
Coefficient variation of service times for each class
E(WBE), E(WLBE)
Average waiting time for each class
ρ
V(SBE), V(SLBE)
E(RBE), E(RLBE)
E(W)
E(R)
E(RS)
K
I
Cumulative utilization
Variance of service times for each class
Average system time for each class
Average global waiting time
Average global system time
Mean residual service time
Percentage impact of Less than Best Effort traffic on
Best Effort traffic
Temporal impact of Less than Best Effort traffic on
Best Effort traffic
80 Chapter 5: Widening Internet Accessibility through Broadband Sharing
For this purpose, several bimodal and trimodal packet-size distributions found in
the literature [118, 119] are considered in the analysis. A list of the examined
packet-size distributions is presented in Table 5-1.
The numerical analysis is carried out in steady-state condition (ρ<1) and is
completed in two stages. First, the behavior of the system when it operates under a
First Come First Served policy (FCFS) with no priorities is analyzed. This part of the
analysis allows for a rough estimation of the average global system time. In case the
range of average global system time values exceeds an acceptable level the access
point should apply proper scheduling mechanisms in order to guarantee that the
impact in terms of additional queueing delay is statistically zero for the home-group
users. Furthermore, the applied scheduling mechanism should also allow guest users
to fully exploit the capacity of the broadband link in case of zero home-user traffic.
In that context, the second stage of the analysis includes the investigation of
system’s behavior under a non-preemptive priority queueing policy with BE-class
traffic having full priority over LBE-class traffic. The notation used throughout the
analysis is summarized in Table 5-2.
5.2.1.1.
FCFS policy (no priorities)
The total arrival rate of the system equals to λ = λBE + λLBE, therefore, the average
global waiting time equals to:
𝐸𝐸(𝑅𝑅𝑅𝑅)
𝐸𝐸(𝑊𝑊) = (1−𝜌𝜌)
Mean residual time for ΣΜ/G/1 systems is given by:
Since 𝐶𝐶𝐶𝐶 2 (𝑆𝑆) =
𝑉𝑉(𝑆𝑆)
𝐸𝐸 2 (𝑆𝑆)
𝛦𝛦(𝑅𝑅𝑅𝑅) =
𝜆𝜆∗𝛦𝛦(𝑆𝑆 2 )
2
(5-1)
(5-2)
and E(𝑆𝑆 2 ) = 𝑉𝑉(𝑠𝑠) + 𝐸𝐸 2 (𝑆𝑆) , E(𝑆𝑆 2 ) can be substituted with
𝐸𝐸 2 (𝑆𝑆) ∗ (1 + 𝐶𝐶𝐶𝐶 2 (𝑆𝑆)) and based on Equations (5-1) and (5-2), the average global
waiting time is produced:
𝐸𝐸(𝑊𝑊) =
𝜆𝜆∗𝐸𝐸 2 (𝑆𝑆)∗(1+𝐶𝐶𝐶𝐶 2 (𝑆𝑆))
2∗(1−𝜆𝜆∗𝛦𝛦(𝑆𝑆))
(5-3)
Towards a Future Internet Architecture: Protocols and Support Mechanisms 81
The average global system time is given by:
5.2.1.2.
𝐸𝐸(𝑅𝑅) = 𝐸𝐸(𝑊𝑊) + 𝐸𝐸(𝑆𝑆)
Strict non-preemptive priority scheduling
(5-4)
The average waiting time for BE- and LBE-class traffic, respectively, equals to:
𝐸𝐸(𝑅𝑅𝑅𝑅)
𝐸𝐸(𝑊𝑊BE ) = (1−𝜌𝜌
BE )
𝐸𝐸(𝑊𝑊LBE ) = (1−𝜌𝜌
(5-5)
𝐸𝐸(𝑅𝑅𝑅𝑅)
BE )(1−𝜌𝜌BE −𝜌𝜌LBE )
Then, the mean residual time is given by:
𝛦𝛦(𝑅𝑅𝑅𝑅) = ∑𝑍𝑍𝑖𝑖=BE
𝜆𝜆𝑖𝑖 ∗𝛦𝛦(𝑆𝑆2𝑖𝑖)
(5-6)
(5-7), where Z= {BE, LBE}
2
Since packet-size distribution is the same for both classes:
𝐸𝐸(𝑆𝑆) = 𝐸𝐸(𝑆𝑆BE ) = 𝐸𝐸(𝑆𝑆LBE )
𝐶𝐶𝐶𝐶 2 (𝑆𝑆) = 𝐶𝐶𝐶𝐶 2 (𝑆𝑆BE ) = 𝐶𝐶𝐶𝐶 2 (𝑆𝑆LBE )
(5-8)
(5-9)
Based on Equations (5-5), (5-6), (5-7), (5-8) and (5-9) the average waiting time for
both classes is calculated as follows:
𝐸𝐸(𝑊𝑊BE ) =
(𝑊𝑊LBE ) =
(𝜆𝜆BE +𝜆𝜆LBE )∗𝐸𝐸 2 (𝑆𝑆)∗(1+𝐶𝐶𝐶𝐶 2 (𝑆𝑆))
(5-10)
2∗(1−𝜆𝜆BE ∗𝛦𝛦(𝑆𝑆))
(𝜆𝜆BE +𝜆𝜆LBE )∗𝐸𝐸 2 (𝑆𝑆)∗(1+𝐶𝐶𝐶𝐶 2 (𝑆𝑆))
2∗(1−𝜆𝜆BE ∗𝛦𝛦(𝑆𝑆))(1−𝜆𝜆BE ∗𝛦𝛦(𝑆𝑆)−𝜆𝜆LBE ∗𝛦𝛦(𝑆𝑆))
(5-11)
Average service times for each class and the average global system time,
respectively, are given by:
𝐸𝐸(𝑅𝑅BE ) = 𝐸𝐸(𝑊𝑊BE ) + 𝐸𝐸(𝑆𝑆)
𝐸𝐸(𝑅𝑅LBE ) = 𝐸𝐸(𝑊𝑊LBE ) + 𝐸𝐸(𝑆𝑆)
𝐸𝐸(𝑅𝑅) =
𝜆𝜆BE
𝜆𝜆
𝐸𝐸(𝑅𝑅BE ) +
𝜆𝜆LBE
𝜆𝜆
(5-12)
(5-13)
𝐸𝐸(𝑅𝑅LBE ) (5-14)
In order to calculate the impact of LBE-class traffic on BE-class traffic, the average
system time of BE class is calculated first in case of zero LBE traffic:
𝐸𝐸′(𝑅𝑅BE ) = 𝐸𝐸′(𝑊𝑊BE ) + 𝐸𝐸(𝑆𝑆), where λLBE=0
82 Chapter 5: Widening Internet Accessibility through Broadband Sharing
Based on Equation (5-10):
𝐸𝐸′(𝑅𝑅BE ) =
𝜆𝜆BE ∗𝐸𝐸 2 (𝑆𝑆)∗(1+𝐶𝐶𝐶𝐶 2 (𝑆𝑆))
2∗(1−𝜆𝜆BE ∗𝛦𝛦(𝑆𝑆))
+ 𝐸𝐸(𝑆𝑆)
Therefore, the average impact in milliseconds (I) equals to:
I = 𝐸𝐸(𝑅𝑅BE ) - 𝐸𝐸′(𝑅𝑅BE )
Percentage-wise the impact can be calculated as:
E(RBE )
E′(RBE )
=1+
λLBE ∗E(S)∗(1+Cv2 (S))
λBE ∗Ε(S)∗(Cv2 (S)−1)+2
(5-15)
(5-16)
(5-17)
From Equation (5-17), a K percentage increase on the average system time of BE
class is observed that equals to:
K=
5.3.
λLBE ∗E(S)∗(1+Cv2 (S))
λBE ∗Ε(S)∗(Cv2 (S)−1)+2
(5-18)
Towards A Hybrid Packet Scheduling Scheme
5.3.1. Impact of Different Packet-size Distributions and Bandwidth Capacities
on Average Global System Time
The results presented in Figure 5-1 are obtained using Equation (5-4). It is clear that
the longest queueing delays are produced by the bimodal packet-size distribution,
while the 58-33-9 trimodal packet-size distribution produces the shortest queueing
delays.
The other two packet-size distribution models present statistically similar behavior,
exhibiting performance close to bimodal distribution. As expected, bandwidth
capacity significantly affects queueing delays, especially under high channel
utilization.
An
interesting
observation
is
that
even
in
high-bandwidth
configurations, average global system time remains close to the bound of 1 second; a
rather significant delay especially for delay-sensitive applications. This fact
constitutes a strong indication that time-sensitive applications ran by the home-
group users might suffer intolerable delays and, hence, calls for a queueing scheme
with priorities that classify packets according to some predefined level of service.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 83
Figure 5-1. Average system time for various packet-size distributions and bandwidth
capacities.
5.3.2. Impact of Guest-user Traffic on Home-user Traffic
The rest of the results presented in the following sections are focused on the
bimodal distribution, in which the system exhibited the worst performance. It
should be noted though that similar analysis could apply to other packet-size
distributions, as well.
Figure 5-2. Average system time for both classes under several combinations of channel
utilization.
84 Chapter 5: Widening Internet Accessibility through Broadband Sharing
Figure 5-2 presents the average system time for both classes under several
combinations of channel utilization, in a system with 500Kbps available bandwidth.
Priority queueing reduces significantly the average system time of BE-class traffic
(under 50ms in most cases), even for high channel utilizations. Note that, in any
case, maximum BE-class delay (see case “BE-Class , λLBE=0, λBE=λ” in Figure 5-2) is
equal or less than average global system time, which here reaches up to 11 seconds
for high channel utilizations.
Both the percentage and temporal impact of LBE-class traffic on average system
time of BE-class traffic are depicted in Figures 5-3 and 5-4, respectively. Various
loads of total channel utilization for two highly-discrepant broadband link speeds
are examined in order to understand how the system behaves under diverse
network configurations. In particular, Figure 5-3 captures the increasing pace of the
impact of LBE-class traffic on BE-class traffic. Figure 5-4 shows that the additional
average delay imposed to BE-class traffic can reach, in the worst case, up to 11ms; a
rather significant delay considering that typical round-trip times (RTT) between
home users’ Asymmetric Digital Subscriber Line (ADSL) access points and popular
websites are ranging between 60 and 100ms. However, it should also be noted that
even for low-bandwidth links a corresponding area exists (see x-axis 0-10% in
Figure 5-4), in which proper regulation of LBE-class traffic can lead to significantly
shorter delays. Therefore, packet scheduling appears in such cases to be a powerful
tool to control the impact on BE-class traffic.
In the following, two potential options for regulating the arrival rate of LBE traffic
(λLBE) are presented: i) Based on Equation (5-18) and setting K appropriately to
confine λLBE (this is the method used in [116]) and ii) Based on Equation (5-16),
where λLBE is regulated by its actual impact on BE-class traffic in milliseconds. A
careful comparison of Figures 5-3 and 5-4 though, reveals that the percentage
impact values do not correspond to the respective values in milliseconds. Consider
for example, the “99.9% Load” line for the 500Kbps and 6Mbps bandwidth
availability cases in both graphs. Although the percentage behavior is exactly the
same, the respective temporal impact differs by almost 10ms. Therefore, the
Towards a Future Internet Architecture: Protocols and Support Mechanisms 85
percentage-based approach for regulating the rate of LBE traffic can be suboptimal
is some cases, since it does not provide any quantitative characteristics of the
additional imposed delay, such as its order of magnitude. Having said that, it is
argued here that Equation (5-16) is more efficient for regulating λLBE.
Figure 5-3. Percentage impact of LBE-class traffic over BE-class traffic.
Figure 5-4. Temporal impact of LBE-class traffic over BE-class traffic.
Finally, the results presented in Figure 5-5 indicate that a bandwidth availability
point exists over which the impact on average system time for BE-class remains
under a specific target value. For a target impact value of 1ms, this bandwidth
availability point is estimated at 6Mbps, whereas if the tolerated impact by BE-class
86 Chapter 5: Widening Internet Accessibility through Broadband Sharing
traffic was relaxed to 3ms, the corresponding bandwidth availability point drops at
2Mbps. This behavior constitutes a clear indication that over a low-bound value of
bandwidth availability, a certain amount of bandwidth could be exclusively
allocated to LBE-class traffic, without having any major impact on BE-class traffic.
The value of such bound may be adjusted towards the level of home users tolerance.
Figure 5-5. Temporal impact of LBE-class traffic over BE-class traffic on high traffic loads.
5.4.
Hybrid Packet Scheduling Scheme
Departing from these observations, HPSS is introduced as a practical way to
overcome most of the identified deficiencies of currently employed queue-
management policies for broadband sharing. HPSS relies on home users’ queueing
delay tolerance level (HPSS target delay) for its operation. In particular, HPSS target
delay is used as the upper bound of the additional queueing delay imposed to home-
user traffic by guest-user traffic. Based on this target delay value, HPSS determines
analytically the HPSS capacity threshold, which corresponds to a particular
bandwidth availability point. This threshold reflects the point based on which HPSS
selects its modus operandi and employs the most appropriate scheduling policy
between priority queueing and class-based WFQ. Practically, for access networks
with maximum capacity exceeding the determined HPSS capacity threshold, HPSS
employs a class-based WFQ policy, in order to guarantee a minimum service rate for
guest users, as long as the total channel utilization is less than 98%. In case channel
Towards a Future Internet Architecture: Protocols and Support Mechanisms 87
utilization exceeds that limit, the system cancels the CB-WFQ policy, and applies a
non–preemptive PQ policy instead, to guarantee that the impact will not exceed a
tolerable value.
The reasoning behind that strategy derives from the fact that in both 60-40 and 60-
20-20 packet-size distributions the additional delay between applying a class-based
WFQ and a PQ policy increases exponentially as channel utilization increases,
according to the results presented in Figure 5-6. In particular, the additional delay
imposed to the average system time of BE-class traffic remains under tolerable
levels for channel utilizations up to 98%; for higher channel utilizations the impact
grows further. At this point, it should be noted that the average system time per
class for the CB-WFQ policy used for producing the results presented in Figure 5-6 is
calculated using the lower-bound mean delay limit equation for the M/G/WFQ
model presented at [120].
Figure 5-6. Additional delay imposed to the average system time of BE-class traffic after
applying a class-based WFQ policy for various percentages of resources exclusively
allocated to LBE-class traffic. Comparison with an unregulated PQ policy.
The amount of resources allocated to serve guest users in the WFQ-based mode of
operation of HPSS is estimated analytically using the following methodology:
First, an HPSS target delay value should be selected. For example, let us consider
here an HPSS target delay value of 3ms. Based on this HPSS target delay value, the
results presented in Figure 5-5 prompt for an HPSS capacity threshold of 2Mbps,
since over this bandwidth temporal impact of LBE-class traffic over BE-class traffic
remains within the acceptable limit of 3ms. Furthermore, the results presented in
88 Chapter 5: Widening Internet Accessibility through Broadband Sharing
Figure 5-6 suggest that an increase in the allocated resources for LBE-class of 0.5%
per Mbps (e.g. 2Mbps: BWLBE = 0.01*BWtotal, 10Mbps: BWLBE = 0.05*BWtotal etc.)
stabilizes the additional imposed delay around a particularly low value, here 2ms.
Thus, a limited overall impact for BE-class traffic is guaranteed, with a maximum
value that is lower than 5ms, considering the 3ms value of HPSS target delay and the
2ms value of the additional impact. Beyond that level of 0.5% per Mbps, statistical
impact becomes significant. (see Figure 5-6 - 1% per Mbps cases, e.g. 2% - 2Mbps
case). All remaining resources, obviously, are allocated to the service of BE-class.
For access networks whose maximum capacity is below the determined HPSS
capacity threshold, HPSS employs a non-preemptive PQ policy between home- and
guest-user traffic. In this case, unlike UPNQ, HPSS employs a temporal-based packetscheduling algorithm to confine the additional delay imposed on home users’
average queuing delay below the HPSS target delay. In more detail, considering the
example used previously, for access networks whose capacity is equal to or lower
than 2MBps, HPSS applies a non-preemptive PQ policy between home- and guest-
user traffic and confines the additionally imposed delay under 3ms by using
Equation (5-16), as the basis for the employed temporal-based packet-scheduling
algorithm.
Finally, it should be mentioned here that the aforementioned values are not
restrictive; instead, HPSS provides high level of flexibility to adjust the specific
parameters according to the administrator or home user requirements. Therefore,
other variants of HPSS may correspond better to other requirements.
6. Evaluation Methodology
The functional complexity and high diversity of the solutions introduced in the
course of this thesis call for the employment of a wide range of tools and metrics in
order to effectively assess their efficacy. Thus, this chapter elaborates on the overall
methodology followed to evaluate both the efficiency and the effectiveness of the
proposed solutions. In particular, the corresponding evaluation goals are identified
in section 6.1. The datasets employed to represent realistic broadband sharing
networking environments, along with the respective analysis, are presented in
section 6.2. The network scenarios upon which the evaluation process was based
are described in section 6.3. The metrics used for assessing the performance of each
solution are defined in section 6.4. Finally, the tools exploited for data analysis and
network systems simulation are discussed in section 6.5.
6.1.
Evaluation Objectives
The overall objectives of the evaluation study conducted in the course of this thesis
are to assess the efficacy of the proposed solutions in effectively fulfilling their
corresponding roles, as described in chapters 3, 4 and 5, respectively, as well as to
gain further insight on the dynamics of the various Future Internet topics explored
herein. Given the particular research directions and design objectives that drove the
89
90
Chapter 6: Evaluation Methodology
development of each solution, the respective evaluation objectives per-solution are
as follows:
The main evaluation objective regarding Bundle Streaming Service is to
investigate the streaming performance benefits associated with BSS’s hybrid
forwarding functionality, especially in highly stressed networking environments.
Although long communication disruptions due to line-of-sight-blocking, mobility,
coverage holes or poor channel conditions are quite frequent in DTN environments,
BSS’s evaluation focuses primarily on the impact of lengthy signal propagation
delays, as well as transient disruptions. The rationale behind this evaluation
strategy is straightforward: the objective of BSS is not to provide a universal
solution by ensuring successful streaming communication no matter how severe the
disruption is, (note that such a solution may not exist when conditions are
extraordinary), but rather to enhance the streaming data delivery performance
under conditions that stop short of permanent connectivity loss.
In order to evaluate the impact of lengthy signal propagation delays and transient
disruptions in data streaming delivery performance, the experiments conducted in
the course of this evaluation study were divided into two main performance
evaluation sets: i) content-insensitive network performance and ii) multimedia
content delivery performance. In both sets, the performance of BSSP is compared
with the performance of IPN forwarder (ipnfw); IPN forwarder is ION’s default
forwarder for forwarding bundles with endpoint identifiers conforming to the IPN
naming scheme. Evaluation of BSS in comparison to other data-streaming
approaches that include routing and FEC techniques is beyond the scope of this
dissertation because the BSS framework is not antagonistically disposed towards
them; on the contrary, it focuses on enhancing the forwarding procedure, while at
the same time, due to its design philosophy, promotes the application of
complementary mechanisms.
With respect to the proposed asynchronous, fully distributed, LEO-based data
dissemination system, the main evaluation objective is to assess its performance
both for varying number of ground stations with different geographical
Towards a Future Internet Architecture: Protocols and Support Mechanisms 91
distributions and for varying number of low-cost satellites supporting the system.
On the one hand, varying the number of ground stations along with their respective
geographical distribution allows for evaluating the impact of those two system
properties on the overall system’s data delivery ratio and latency. On the other
hand, estimating the total data return volume achieved by the broadcast model
while varying the number of low-cost satellites provides a rough idea of the
comparative performance between dense, low-cost satellite networks versus the
traditional, high-end, point-to-point approach. At this point, it should be noted that
comparing the broadcast peer-to-peer model, supported solely by a single satellite,
against the traditional, high-end Space communication models would be largely
unfair due to the major difference in the associated deployment cost and equipment
specifications. Therefore, such comparison evaluations are considered out of scope
in the context of the present evaluation study.
Finally, the potential widening of Internet access through broadband sharing is
also targeted; it becomes apparent early on that a successful domestic sharing
scheme should efficiently mitigate the set of contradictory requirements developed
between home and guest users, while, at the same time, providing some level of
service to the guest users. Both aspects are taken into consideration in the present
evaluation study through two distinctive evaluation goals. The first evaluation goal
is associated with comprehending the dynamics of a broadband sharing system in a
worst-case-scenario situation, in which guest users request service at a time when
home users make full use of their bandwidth, and explore whether certain queuing
policies and techniques can successfully mitigate the occasionally contradictive set
of requirements developed between home and guest users. This is achieved by
experimentally investigating the impact of guest-user traffic on sharers’ network
performance under various queue-management configurations, including the
proposed Hybrid Packet Scheduling Scheme. Since subscribers’ control over
downlink traffic is limited, the present evaluation study focuses on the uplink traffic,
which can be regulated at gateways placed at the edges of networks, such as
wireless access points. The second evaluation goal is associated with investigating
92
Chapter 6: Evaluation Methodology
whether domestic broadband connection sharing schemes could indeed be used as
frameworks for providing simultaneous free Internet access to a considerable
number of guest users without hampering sharers’ network performance.
Overall, both solution-specific and generic conclusions, which are not limited to the
specific conditions defined by the experimental scenarios described below, can be
drawn based on the evaluation study conducted in the context of this thesis.
Therefore, the obtained results provide both a comprehensive perspective and a
fundamental understanding of the explored issues, which could form a solid basis
for future research.
6.2.
Datasets Analysis for Broadband Sharing Simulations
Building realistic broadband sharing simulation scenarios calls for the accurate
modeling of certain access network properties and guest-user uplink traffic. With
respect to the present evaluation study, these models were produced by analyzing
actual traffic traces and usage data, collected in the context of the PAWS project trial
[121], including data from 17 open-access broadband sharing networks and 15
guest users.
The deployment of PAWS took place over a seven-month period (July 2013 -
February 2014) in Nottingham, UK. 20 PAWS gateways were deployed in total, 8 of
which were used by 15 citizens to freely access the Internet. From the 12 remaining
access points, 9 served exclusively as measurement points, while 3 were switchedoff most of the time.
Each deployed PAWS gateway, depicted in Figure 6-1, carried out daily
measurements of the throughput, latency and loss experienced by each sharer's
broadband network. Those measurements allowed for evaluating the performance
of each access network broadband connection. Measurements were carried out
regardless of whether the sharers’ access point was actually used by some guest
user. The following data were collected by PAWS team members and anonymized
before being uploaded to a central repository:
i) Throughput: Upload and download throughput (Mb/s) was measured every six
Towards a Future Internet Architecture: Protocols and Support Mechanisms 93
hours, using three parallel TCP threads to provide an accurate estimate of access
link capacity.
ii) End-to-End RTT (e2ertt): A set of servers in the UK were pinged and the
corresponding Round-Trip-Times (RTT) measurements (in ms) were collected.
iii) Last-Mile RTT (lmrtt, ulrttdw and ulrttup): lmrtt represents last mile latency (ms)
under normal network load, whereas ulrttdw and ulrttup represent the downstream
and upstream last mile latency (ms), under heavy network load. The ulrttdw and
ulrttup values are specified by measuring last mile latency, while bitrate tests are
on. These values were collected on a 2- hour interval basis.
Finally, guest-user traffic traces from the PAWS network were collected on the VPN
server, also depicted in Figure 6-1. During the deployment a total of 29GB traffic
(10GB upload, 19GB download) was generated by 15 citizens.
Figure 6-1. Public Access Wi-Fi architecture.
6.2.1. Modeling Access Points
The downlink and uplink capacity of an access network, along with the two
corresponding queue sizes, constitute the four most essential properties
characterizing an access network profile. In order to model these four properties,
data collected during the PAWS project trial from 17 access networks (AN) were
analyzed. Statistical measurements such as mean, median and inter-quartile range
94
Chapter 6: Evaluation Methodology
(IQR) are used throughout the analysis to identify the salient characteristics of the
sample data.
Each one of the 17 access network profiles is abbreviated in Figure 6-2 by AN,
followed by a corresponding identification number. More specifically, Figure 6-2(a)
shows the downlink capacity of each access network, along with its corresponding
downlink queue-size values range, whereas Figure 6-2(b) shows the respective
values for the upstream direction.
(a)
(b)
Figure 6-2. Access network profiles produced from the analysis of data collected during the
PAWS project trial. (a) Downlink capacity of each access network along with its
corresponding downlink queue-size values range (b) Uplink capacity of each access
network along with its corresponding uplink queue-size values range. Boxes show
interquartile ranges and the median, with the whiskers showing the 10th and 90th
percentiles.
6.2.1.1.
Access network capacity
The capacity for each access network is determined based on the reported
throughput values. In order to capture the actual uplink and downlink capacity of
each access network, independently of the throughput variations throughout
different time periods, the average value of the respective reported values is used.
The network capacity results presented in Figures 6-2(a) and 6-2(b), respectively,
for the downstream (Bandwidthdw) and upstream (Bandwidthup) direction, are
used for the rest of the analysis.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 95
6.2.1.2.
Access network uplink and downlink queue size
Access point’s downlink and uplink queue-size values for each access network are
determined based on lmrtt, ulrttdw and ulrttup values. In particular, the buffering
effect is quantified, in kilobytes, for each traffic direction as follows:
𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑑𝑑𝑑𝑑 =
𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑢𝑢𝑢𝑢 =
(𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢−𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙)∗𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵ℎ𝑑𝑑𝑑𝑑
8
(𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢−𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙)∗𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵ℎ𝑢𝑢𝑢𝑢
8
(6-1)
(6-2)
For the first five access networks, which employ fiber broadband links, most of the
observed downlink median queue-size values are close to 1.2MB, whereas the
respective uplink median queue-size values are between 60 and 100KB. Regarding
the rest of the access networks, results for the downstream direction can be divided
into two sub-categories. Access points that employ high-speed conventional ADSL
links (i.e. 20Mbps) present downlink median queue-size values close to 1MB,
whereas, most of the access points that employ low-speed ADSL links (i.e. 3 –
12.5Mbps) present downlink median queue-size values close to 75KB. The
respective uplink median queue-size values for all access points employing
conventional ADSL links are approximately 50KB.
Although most of the results for both downstream and upstream direction are
consistent, there are cases where the median queue-size values diverge significantly
from the respective median values of access networks with similar network capacity
profiles (e.g. AN10, AN11, AN13 and AN17). This fact could be attributed either to a
tweaked access point configuration or to some error introduced by the tool used to
take the lmrtt, ulrttdw and ulrttup measurements. Finally, some slight variations
from the aforementioned median queue-size values are observed across all cases,
which is considered reasonable due to the high diversity in the network
characteristics of the investigated access networks.
6.2.2. Modeling Guest-user Traffic
The employment of realistic network traffic patterns constitutes an essential
simulation aspect that significantly contributes to the overall accuracy of
96
Chapter 6: Evaluation Methodology
simulation-based studies. In the context of the present evaluation study, realistic
uplink guest-user traffic patterns are employed to drive the broadband sharing
simulations.
In order to produce those patterns, the collection of guest-user traffic traces,
captured during the PAWS project trial, are first analyzed by following a flow-level
analysis approach. To describe the properties of the investigated traffic sufficiently,
the following flow-level characteristics are considered: i) flow inter-arrival time –
the time interval between two consecutive flow arrivals, measured in seconds, ii)
flow size – the total number of bytes transferred during a TCP flow and iii) flow
duration – the time between the start and the end of a TCP flow, measured in
seconds.
After extracting the samples of the aforementioned flow-level characteristics from
the guest-user traffic traces, those samples are then used to derive statistical models
that are capable of producing similar uplink guest-user traffic patterns to those
observed during the trial. In order to accomplish that, both goodness-of-fit tests and
Probability-Probability (P-P) plots are used to assess the adequacy of various
statistical distributions against the empirical ones.
6.2.2.1.
The broader picture
Out of the 15 citizens that took part in the trial as guest users, four of them were
responsible for the largest part of the traffic, with one of them – the primary citizen –
being responsible for over 75% of the total guest-user traffic [121]. Due to the
limited amount of data collected for the rest of the guest users, our modeling effort
was focused on those four individuals. For each one of the four guest users, the three
corresponding empirical distributions (i.e. inter-arrival time, flow size and flow
duration
empirical
distributions)
produced
by
the
respective
flow-level
characteristic samples was attempted to be optimally fitted to a wide range of
typical statistical distributions. Three different statistical tests were employed to
evaluate goodness-of-fit in each of these cases, namely the Kolmogorov-Smirnov,
Anderson-Darling and Chi-Squared tests. The majority of these tests returned
Towards a Future Internet Architecture: Protocols and Support Mechanisms 97
1
negative results. Various data transformations, such as √𝑥𝑥 and , and combinations
𝑥𝑥
of statistical distributions were also employed during the modeling effort, achieving
poor to moderate results. In order to compensate for these discrepancies, the
goodness-of-fit of various statistical distributions was empirically evaluated on the
sample data.
6.2.2.2.
Empirical analysis
P-P plots are employed to empirically assess the goodness-of-fit of various statistical
distributions on the sample data. In particular, goodness-of-fit is determined by
manually inspecting the overall deviation of the statistical distribution plot from the
empirical distribution line; the statistical distribution presenting the smallest
deviation is chosen as the best-fit distribution. Each of the three panel rows of
Figure 6-3, presents the P-P plots for inter-arrival times, flow sizes and flow
durations, respectively. The first column of panels, in particular, panels (a), (c) and
(e), present the P-P plots regarding the primary citizen. In order to have a good
grasp of the traffic properties at both ends of the four investigated cases, the second
column of panels present the P-P plots regarding the guest user who generated the
least amount of traffic. The respective P-P plots for the other two guest users, which
were also produced in the course of the empirical analysis, present similar behavior
to the ones depicted in Figure 6-3. For clarity reasons, each panel depicts the
empirical distribution line, along with the plots of the five statistical distributions
that present the smallest overall deviation.
1) Flow Inter-arrival times
In Figures 6-3(a) and 6-3(b), the empirical distribution of inter-arrival times is
compared with several statistical distributions. In both cases, Weibull distribution
presents the best fit, showing only minimal deviation from the empirical
distribution line. This is in accordance with recent studies [122] [123], which also
employ Weibull distribution for modeling TCP flows inter-arrival times in
networking environments similar to the one investigated here.
98
Chapter 6: Evaluation Methodology
More specifically, in Figure 6-3(a), although both Log-logistic and Weibull
distributions present similar overall behavior, Weibull seems to have a slight
advantage over Log-logistic distribution at the left tail. In particular, there are many
points of the Log-logistic distribution in the range of 0 to 0.2, which present
considerably larger deviation from the empirical line over the respective Weibull
distribution points. In Figure 6-3(b), Burr, Lognormal and Weibull distributions
present similar behavior, with Lognormal and Burr distributions having larger
deviation over the empirical distribution line at the left tail. The rest of the
distributions presented in Figures 6-3(a) and 6-3(b) are, evidently, not appropriate
matches.
2) Flow sizes
It is clear from Figure 6-3(c) that Generalized Logistic and Generalized Pareto
distributions present almost identical behavior with minor deviations from the
empirical distribution line, making them both possible candidates for modeling flow
sizes. The results depicted in Figure 6-3(d) show that all five distributions present
similar light-tailed behavior. The same results also suggest that Lognormal and
Generalized Pareto distributions perform slightly better in comparison with the
other distributions. Generalized Pareto is the common denominator in both cases;
therefore, it is selected as the preferred statistical distribution for modeling flow
sizes. Pareto-type distributions have also been used in previous studies for modeling
TCP flow sizes [124][125].
3) Flow durations
The plots presented in Figures 6-3(e) and 6-3(f) clearly demonstrate the inability of
the presented distributions to sufficiently match the respective flow duration
empirical distribution. Lognormal distribution seems to perform slightly better in
comparison with the rest of the examined distributions. This behavior remains
consistent across all four examined cases. Due to this fact, Lognormal is selected as
the preferred distribution for modeling flow durations in the present evaluation
study.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 99
(a)
(b)
(c)
(d)
(e)
(f)
Figure 6-3. P-P plots comparing the empirical distribution of inter-arrival time, flow size
and flow duration characteristics with various statistical distributions.
100
Chapter 6: Evaluation Methodology
6.2.2.3.
Experimental validation of empirical analysis results
Both analytical and empirical goodness-of-fit results suggest that none of the
selected statistical distributions for modeling inter-arrival times, flow sizes and flow
durations constitute a perfect fit to the respective sample data. In order to validate
the suitability of the selected distributions set and estimate potential fitting errors,
the uplink traffic properties generated by the empirical versus the analytical model
were experimentally compared, for all four guest-user profiles.
For this purpose, a flow-based traffic generator was developed in NS2. In the
empirical modeling case, inter-arrival times, flow sizes and flow durations are
sampled from the respective empirical distributions, whereas in the analytical
modeling case, the same flow-level characteristic values are sampled, respectively,
from the Weibull, Generalized Pareto and Lognormal distribution.
The following metrics are considered to validate the appropriateness of the selected
distributions set in modeling the uplink guest-user traffic patterns observed during
the PAWS project trial: i) total number of generated packets, ii) total number of
received packets, iii) average throughput, iv) max throughput (to capture network
traffic burstiness), and v) average end-to-end delay.
A simple network topology consisting of two nodes is used for the comparison
experiments, with the first node being the source and the second one the
destination. A Droptail queue is also attached to the source node. Simulation time is
set to 1 hour and 100 simulation runs are completed per modeling case and per
guest-user profile.
After slight adjustments to the distributions parameters estimated during the
empirical analysis phase, the selected guest-user profiles presented in Table 6-1
yield average fitting errors of less than 5%, for all metrics. The respective fitting
errors per metric and per guest-user profile are presented in Table 6-2.
In conclusion, the analysis results suggest that the vast majority of the captured
guest-user traffic is composed by mice flows. This fact is clearly indicated by the
Generalized Pareto distribution parameter values. Furthermore, the average
Towards a Future Internet Architecture: Protocols and Support Mechanisms 101
throughput values produced by the guest-user profiles presented in Table 6-1 are in
the range of 1 to 5 KBps, with peak values in the range of 5 to 30 KBps.
Guest-user
Profile
1
2
3
4
a
Table 6-1. User Profile Distribution Parameters.
Flow Characteristics
Inter-arrival time
Parameters a
α = 0.27, β = 0.4
Size
κ = 0.59, σ = 544, μ = 353
Inter-arrival time
α = 0.31, β = 0.53
Duration
σ = 2.62, μ = 1.03
Size
κ = 0.77, σ = 1108, μ = 203
Inter-arrival time
α = 0.4, β = 0.19
Duration
σ = 1.83, μ = 1
Size
κ = 0.12, σ = 1473, μ = 471
Inter-arrival time
α = 0.38, β = 9.58
Duration
Size
Duration
σ = 1.18, μ = 0.9
κ = 0.59, σ = 620, μ = 42
σ = 2.45, μ = 1.97
The presented values refer to the statistical distribution parameters used for
modeling the respective flow characteristic (i.e. Weibul for inter-arrival times, where
α and β are the respective shape and scale parameters, Generalized Pareto for flow
sizes, where κ, σ and μ are the respective shape, scale and location parameters, and
Lognormal for flow durations, where μ and σ are the respective mean and standard
deviation distribution parameters.)
102
Chapter 6: Evaluation Methodology
Table 6-2. Average Fitting Error Between Empirical and Analytical Model.
User Profile
Metrics
2
3
4
Generated Packets
1.62%
0.82%
4.99%
0.54%
Max throughput
3.15%
0.13%
4.54%
3.08%
Received Packets
Avg. Throughput
E2E Delay
6.3.
1
1.62%
1.75%
0.27%
0.82%
0.82%
0.01%
4.99%
5%
0.12%
0.54%
1.15%
0.1%
Scenarios
A set of experimental scenarios was developed as the means to assess the efficacy of
the network solutions introduced in the course of this thesis. Each of these scenarios
is specifically tailored to evaluate the different architectural, performance and
efficiency aspects of a particular solution among the proposed ones with respect to
the evaluation goals described in section 6.1. In general, the developed scenarios
cover different use cases within the general context of Future Internet networking
and are classified into three experimental categories correspondingly to the three
proposed solutions.
6.3.1. Experimental Category I – Data Streaming
Bundle Streaming Service was proposed in the context of this thesis as a framework
for improving DTN architecture streaming capabilities; thus, all scenarios described
in this subsection focus on exploring the performance benefits associated with BSS’s
hybrid forwarding functionality.
Although BSS could be a useful tool for various use cases with notable impact on the
user’s viewing experience, the streaming of data from space to Earth is the most
appropriate use case to demonstrate its potential. This is justified primarily by the
specific application and networking requirements of Space communications that
frequently suffer severe network delays and disruptions. However, BSS could also
Towards a Future Internet Architecture: Protocols and Support Mechanisms 103
demonstrate its potential in highly-stressed terrestrial networking environments,
such as those where tactical military communications are deployed [126].
In order to take into account both of these aspects, a set of flexible scenarios, which
are based on realistic use cases, was developed to evaluate BSS performance under
various communication patterns. These scenarios were built upon a diverse sample
of stressed networking environments, including both terrestrial and Space network
topologies, which incorporate communication links with varying properties.
(a)
(b)
(c)
Figure 6-4. All three network topologies employed in BSS-specific Space scenarios. a)
Network topology employed in the Moon scenario and its associated properties, b) Network
topology employed in Spacecraft scenario and its associated properties and c) Network
topology employed in Mars scenario and its associated properties.
Scenario 1
The first and most common scenario involves streaming data between a Space asset
in close proximity to Earth and a mission operation center on Earth. The topology
comprises a rover on the Moon, a relay satellite that is orbiting Moon and a ground
station on Earth (see Figure 6-4(a)). Given the fact that the round-trip
communication delay between Earth and Moon is less than 3s, the propagation
delay for the Moon-relay communication link is set at 500ms and for the relay-Earth
communication link at 1s. For this scenario, BER is set to 10-6 for every
104
Chapter 6: Evaluation Methodology
communication link based on the short distance between the participating objects in
terms of Space communication ranges.
Scenario 2
In this scenario, a deep Space communication use-case is considered where a distant
spacecraft sends a stream of data to the mission operation center on Earth through
an intermediate Space asset, which could be another spacecraft or space station (see
Figure 6-4(b)). The calculation of the minimum time that a signal needs to travel a
deep Space distance1 results in a rough value of 7s. In this scenario, an arbitrary
propagation delay value of 20s per link was set to emulate networking
environments with propagation delays well beyond 7s. In terms of communication
links BER, this scenario is more complicated than the first one. The low-power
antennas of the spacecraft justify a BER value of 10-5 for the communication link
between the two spacecraft, while a 10-6 BER value is more appropriate for the
communication link between Earth and the intermediate spacecraft, given the
increased capabilities of the equipment installed on Earth. To better understand this
reasoning, one must keep in mind that the deep Space network is equipped with
massive antenna arrays that are able to enhance the reception of low power/weak
signals.
Scenario 3
This last BSS-specific Space scenario reflects the topology and the operations of the
ongoing exploration missions on Mars. A rover on Mars streams data to a ground
station on Earth via a relay satellite orbiting Mars (Figure 6-4(c)). The propagation
delay for the long-haul links between Mars and Earth varies from 3 to 20min,
because the distance between the two planets varies as they move in their orbits. In
order to have a reasonable emulation time interval that allows adequate experiment
repetitions to be conducted, a total average end-to-end propagation delay of 5
minutes that remains stable for the total duration of the data streaming session is
assumed for the specific scenario. In this scenario, which is related to interplanetary
1
According to the official definition given by ITU, Deep Space starts at a distance of 2,000,000 km from the Earth's surface.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 105
communications, BER between the rover landed on Mars and the relay satellite is
set at the value of 10-6, which is appropriate given the relatively short distance
between these two objects. Because of the long-haul distance between the relay
satellite and Earth, the BER for the communication link connecting the two entities
is set to 10-5.
At this point, it should be noted that the BER values used in the aforementioned
scenarios were extracted from descriptions of communication links found in Space
communications literature. More specifically, the most frequently quoted BER
values are in the order of 10-5 and 10-6 [127]; therefore, BER values in between this
range were applied. In that context, the selection of the proper BER value for each
communication link is mainly influenced by the distance that a signal has to travel as
well as the specific technological properties of the communication equipment used
in each case. It would also be useful to note here that the applied BER values refer to
errors after applying error correction codes and hence reflect the BER values
experienced by the bundle layer, the rate at which uncorrectable bit errors are
detected. The ultimate aim is to emulate the properties of actual space
communication topologies with a sufficient level of accuracy, while at the same time
retaining a high level of flexibility in the scenario development and configuration
process.
Figure 6-5. All three network properties classes of the topology employed in BSS-specific
terrestrial scenarios.
Scenario 4
Scenario 4 refers to the deployment of BSS in highly-stressed terrestrial networking
environments. Of course, it is impossible for the developed terrestrial scenario to
106
Chapter 6: Evaluation Methodology
capture the wide diversity of parameters, which can be encountered in various
networking terrestrial environments. Instead, the extensive Space scenarios
presented above are coupled with an abstract terrestrial scenario consisting of a
generic, symmetric, 3-node string topology (Figure 6-5), in order to cover a fair
percentage of use cases. In this context, three distinctive sub-scenarios are derived
from the abstract terrestrial scenario depending on the magnitude of propagation
delay alone. Each sub-scenario is named accordingly as follows: (i) normal, which
includes propagation delays of few milliseconds that are typical in local area
networks topologies; (ii) low, which includes propagation delays in the order of tens
of milliseconds, typical in well-interconnected networks; and (iii) high, which
includes propagation delays in the order of hundreds of milliseconds, encountered
usually in highly-stressed networking environments such as the ones established in
battlefields and civil emergency situations. In each case, it is assumed that BER
remains stable at 10-5 as a result of transmission rate adaptation mechanisms
employed by IEEE 802.11 protocol to provide a steady minimum transmission rate,
sufficient for multimedia streaming.
Scenario 5
Scenario 5 refers to the performance evaluation of BSS regarding delivery of
multimedia content. In this scenario, a sending entity streams video to a receiving
entity for 2 minutes. With respect to the employed network topology, the main
requirement was to utilize a network environment that allow for capturing severely
distorted real-time display images, whose quality can be easily assessed even by the
human eye, in order to strongly showcase BSS’s advantage over multimedia content
delivery applications. Among the network topologies presented in this section so
far, the only one conforming to this requirement was the one that was employed in
the Moon scenario, and, subsequently, in this scenario as well.
6.3.1.1.
Reference scenarios
The scenario design phase regarding this experimental category was completed
with the development of reference scenarios to compare the base performance of
the two data streaming settings (default versus BSS-enabled) and to assess the
Towards a Future Internet Architecture: Protocols and Support Mechanisms 107
progressive impact of various conditions on the streaming procedure as well. For
each one of the scenarios described earlier, the applicable corresponding scenarios
with zero BER were used as reference scenarios.
6.3.1.2.
Scenarios variations
In order to evaluate the impact of the number of hops per se on BSS performance,
three additional network topologies were considered, each consisting of five nodes;
the end-to-end propagation delay of the 5-node network topologies remains the
same as for the corresponding 3-node topologies, while BERs are distributed as
depicted in Figure 6-6.
Furthermore, in the context of assessing the impact of different bundle sizes on the
performance of the two forwarding approaches, two different payload sizes were
considered: the commonly used value of 1316 bytes and its double, 2632 bytes.
(a)
(b)
(c)
Figure 6-6. BER and propagation delay distribution on each 5-node network topology
variation. a) 5-node network topology employed in Spacecraft scenario and its associated
properties b) 5-node network topology employed in Mars scenario and its associated
properties and c) All three network property classes of the 5-node network topology
variation employed in terrestrial scenarios.
108
Chapter 6: Evaluation Methodology
6.3.2. Experimental
Category
II
–
Asynchronous,
LEO-based,
Data
Dissemination Model
In order to assess the efficacy of the proposed asynchronous data dissemination
model, a basic content distribution scenario is considered. In this scenario, content
distribution is performed by employing a broadcast communication pattern, where
low-cost satellites orbiting the Earth in LEO orbits broadcast content to small-scale
ground stations, which then deliver the content to the end-users. Since the focus of
the evaluation conducted in this thesis regarding the performance of the proposed
asynchronous data dissemination model is on the Space segment, a single end-user
is actually employed. Figure 6-7 depicts an abstract representation of the topology
employed in this category of experiments.
Figure 6-7. Content distribution scenario network topology.
Content is assumed to be uploaded to the satellites at regular intervals from
dedicated ground stations; therefore, uplink capability is assumed to only be
available to selected entities. The content itself might be composed by several types
of information including news feed, emergency signals, educational material etc.
6.3.2.1.
Scenario variations
The low-cost broadcast communication pattern has been applied to various
simulation topologies with one or more satellites and a dense network of ground
Towards a Future Internet Architecture: Protocols and Support Mechanisms 109
stations. Namely, the employed topologies contain 1 satellite with 52, 100 and 140
ground stations and 6 and 46 satellites with 52 ground stations.
6.3.2.2.
Selection of satellite orbits, satellites bandwidth, and ground stations
locations
The employed low-cost broadcast communication pattern yields a large number of
short, low-speed, and possibly overlapping contacts between satellites and ground
stations. Therefore, the selection of satellite orbits in combination with the
corresponding ground stations locations should be optimal in order to maximize the
efficiency of the system. Regarding the scenario variations described in the previous
subsection, the selection of the most appropriate satellite orbits and ground stations
locations for each case was performed through a separate evaluation study. This
particular study was conducted in cooperation with researchers from Telespazio
VEGA UK, in the context of “Application of a BitTorrent-like Data Distribution Model
to Mission Operations” project, funded by ESA, and was based on STK: a physical
simulator that is capable of modeling satellite orbits along with the respective link
budgets given certain ground stations locations.
First, ground stations locations were selected based on existing templates of
currently deployed ground stations. The locations of these ground stations are
spread across the globe, including areas and places such as Europe, Canada,
overseas territories, world-wide embassies and ESA-operated ground stations
locations.
Then, a selection of transmission systems based on UHF-, S- and X-Bands were
compared so that the most efficient system for the low-cost broadcast model would
be found, in terms of contact opportunities duration, communication link quality
and power consumption. The link-up and link-down events for each of these
contacts are defined as when the calculated BER exceeds a configurable threshold
parameter, typically set at 10-6 for the simulations conducted in the context of this
thesis.
The results indicated that S-band is the most appropriate solution, since it allows for
longer uninterrupted contacts, increased total contact time and decreased BER at
110
Chapter 6: Evaluation Methodology
system level. The calculated BER corresponds to the effective error after FEC has
been applied on the physical link. The physical parameters finally selected for
modelling the space link are S-Band 3 Mbit/s, BER 10-6). In all cases, satellites are
orbiting in sun-synchronous orbit at 700 Km, with an inclination of 96.7 degrees.
The sun-synchronous orbit is selected here because it provides global coverage of
the Earth’s surface, resulting in a higher number of contact opportunities.
Finally, Table 6-3 presents contact opportunities statistics for each one of the
examined scenario variations, by considering a simulation time of 10 days, which is
the average time required for a sun-synchronously orbiting LEO satellite to cover
the entire globe. Avg. Duration is the average contact duration, and Total Duration
and Total Capacity are the contact duration and capacity after any overlapping
contacts have been merged.
Link Model
Table 6-3. Contact Opportunities Statistics.
52 GSs, 1 Satellite
100 GSs, 1 Satellite
140 GSs, 1 Satellite
52 GSs, 6 Satellites
52 GSs , 46 Satellites
6.3.2.3.
Avg. Duration
Total Duration
Total Capacity
7.4 mins
3.1 days
100 GB
7.4 mins
7.4 mins
6.6 mins
5.8 mins
1.3 days
3.7 days
8.1 days
40.7 days
43.4 GB
119 GB
260 GB
1.3 TB
Content upload and data expiration time
For the scenario considered in this category of experiments, it is assumed that data
is uploaded to the satellites at regular intervals by following a uniform upload
pattern. The total amount of data uploaded on each satellite per day is set at 2GB,
totaling 20GB per satellite for the whole simulation time of 10 days. This amount of
data is considered to be sufficient for distributing important news in multimedia
format along with other types of material, such as educational content and
emergency signals.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 111
Finally, given that a satellite orbiting at 700km needs about 100 minutes to
complete a single orbit, a data time-to-live (TTL) value of 12 hours is used for the
experiments conducted in the context of this experimental category in order to
guarantee that the satellite has completed a substantial number of orbits before
data is discarded.
6.3.3. Experimental Category III – Broadband Sharing
In order to explore the effectiveness of various broadband sharing configurations in
sharing domestic broadband network resources to guest users the following
networking scenario is considered:
A family of four people (i.e. home users) is sharing its broadband connection. Each
home user is performing a different task online: home-user 1 is uploading some
photos to the cloud, home-user 2 is holding a Skype video call, home-user 3 is playing
an online game and home-user 4 is browsing the Internet. At the same time, an
arbitrary number of guest users can connect to the family access point and request
various web services.
Figure 6-8. Network topology considered for broadband sharing simulations. Three
different types of links are employed. Wireless and high-bandwidth links have the same
properties across all simulations, whereas the link representing access point’s connectivity
to the Internet has different properties according to the respective access network profile
used in each simulation.
112
Chapter 6: Evaluation Methodology
The topology depicted in Figure 6-8 is used in the simulations to represent the
aforementioned scenario. As shown in Figure 6-8, home and guest users are
connected through 100Mbps wireless IEEE 802.11n links to the access point. The
access point is connected to the Internet through a conventional ADSL or fiber link.
6.3.3.1.
Scenario variations
As shown in Figure 6-2, access points and connectivity links with diverse
characteristics are employed in practice by home users. In order to capture this
diversity and study how different access point configurations and link
characteristics affect the impact of guest-user traffic on home users’ network
performance, 4 different access network profiles are considered in the simulations.
The first two access network profiles are supported by fiber connectivity links (see
Figure 6-2: AN1 and AN3), whereas the other two by conventional ADSL links (see
Figure 6-2: AN8 and AN16). The link properties are defined according to the
respective access network profiles used in each simulation, while access point
queue sizes are set to a value that is equal to the corresponding 90th percentile
values presented in Figure 6-2.
An increasing number of guest users (i.e. 0-30) is also being simulated, in order to
investigate the scalability of the respective access network and queue-management
configurations. All four identified guest-user profiles are considered; their selection
is performed circularly. In order to guarantee that the access network connectivity
link to the Internet constitutes the sole bottleneck point in the network, each
application is served by a different high-bandwidth network server. All applications
remain active throughout the whole simulation time, which is set at 10 minutes.
6.3.3.2.
Queue-management configurations
Seven different queue-management configurations are considered, in total. These
configurations employ combinations of widely available dropping and scheduling
policies, which cover a broad range of the available queue-management policies
spectrum. Since the main interest is placed on practical solutions, purely
experimental or hard-to-implement policies are refrained from the current study.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 113
Instead, the focus is put either on policies that are currently available by network
routers or on policies that can be easily implemented based on existing router
Quality-of-Service (QoS) frameworks. In particular, the following queuemanagement configurations are considered:
i) FIFO queueing, with a simple drop last dropping policy. This configuration implies
that the access point is simply left unlocked for guest users to use.
ii) FIFO queueing with different AQM dropping policies, including Random Early
Detection (RED) and Controlled Delay (CoDel). In contrast to RED, which
probabilistically drops packets based on certain queue size thresholds (min_thresh,
max_thresh), CoDel checks periodically (interval) the average queueing delay and
probabilistically drops packets when a certain queueing delay threshold is exceeded
(target delay). The application of such dropping policies results in lower queueing
delays and increased system fairness.
iii) Smoothed Round Robin fair queueing (SRR) [128]. SRR is a fair queueing (FQ)
packet scheduler, which is attractive due to its very low time complexity. With this
configuration, scheduling effort is distributed evenly and smoothly across all flows
requesting service.
iv) Strict non-preemptive PQ implemented with two FIFO queues, one for each class
of traffic. Similarly to cases (i), (ii) and (iii), guest users are allowed to request
service from the access point in an unregulated manner, but with this configuration,
home-user traffic has non-preemptive priority over guest-user traffic.
v) Regulated non-preemptive priority queuing as implemented by UPNQ. UPNQ
parameters are selected based on the parameter values suggested by the authors in
[116]. UPNQ threshold represents a queue percentage threshold for home-user
packets, over which all newly arriving guest-user packets are forcefully dropped.
Similarly to (iv), with this configuration, home-user traffic has non-preemptive
priority over guest-user traffic, but, additionally, guest-user traffic arrival rate is
regulated, using the UPNQ percentage-based algorithm.
114
Chapter 6: Evaluation Methodology
vi) Hybrid queueing as implemented by HPSS. HPSS target delay is set at 3ms. This
target delay value is considered as a reasonable impact value that home users can
tolerate. The corresponding HPSS capacity threshold value is 2Mbps. In access
networks AN8 and AN16, which have an upload capacity that is below 2Mbps, HPSS
behaves as a regulated non-preemptive priority queueing scheduler, while in access
networks AN1 and AN3, with an upload capacity that is above 2Mbps, HPSS behaves
as a class-based WFQ scheduler. In the first case, guest-user traffic arrival rate is
regulated, using the HPSS temporal-based algorithm. In the second case, the amount
of resources allocated to serve guest-user traffic for access networks AN1 and AN3
is, respectively, 3.25% and 1.5%. Both the HPSS capacity threshold and the amount
of resources allocated to serve guest-user traffic are determined based on the
analysis results presented in [10].
vii) Class-based Weighted Fair Queueing, implemented with two FIFO queues, one
for each class of traffic. 95% and 5% of system resources are allocated, respectively,
for serving home- and guest-user traffic. This particular configuration guarantees a
low impact on home-user throughput, while it is slightly more aggressive than the
respective configuration employed by HPSS.
6.3.3.3.
Traffic patterns
Home-user traffic is composed of 2 elephant and 2 mice flows. Elephant flows are
employed in this setting to represent the increased use of cloud syncing and video
conferencing applications, both of which produce large amounts of uplink traffic.
Furthermore, although it would be more common for home users to engage in
multiple network activities simultaneously, it is argued here that this particular
home network setting represents a realistic traffic mix. This argument is further
supported by the employment of different types of network traffic, as well as by the
10-minute simulation time. More specifically, an FTP application is used to simulate
the traffic generated from the first home user, a CBR application with 500Kbps
bitrate is used to simulate the Skype video call and an on/off traffic generator
application, with an average bitrate of 12.8Kbps that creates a packet every 50ms, is
used to simulate the updates sent to the game server [129]. Another traffic
Towards a Future Internet Architecture: Protocols and Support Mechanisms 115
generator, which sends new web requests following Lognormal inter-arrival times
[130], each with a size of 350bytes, is used to simulate the traffic generated from the
fourth home user.
Unlike home-user traffic, the analysis presented in subsection 6.2.2 suggests that
guest-user traffic is composed of a large number of mice flows. This type of traffic is
generated through the same traffic generator application used in subsection 6.2.2.3.
6.4.
Metrics
In this section, the evaluation metrics utilized on each experimental category are
presented:
With respect to the experiments conducted in the context of the data streaming
experimental category, both performance and quality metrics were employed in
order to accurately assess the networking and multimedia efficacy of BSS and to
evaluate the impact of different network properties, such as bundle size and hop
count, on BSS performance as well. In particular, the following network
performance metrics are considered: i) stream delivery efficiency (SDE) is defined as
the number of bundles received from the final recipient of the stream in a specific
period, starting from the time that the recipient received its first bundle; ii) stream
delivery attenuation (SDA), as the number of out-of-order bundles received from the
final recipient of the stream in a specific period, starting from the time that the
recipient received its first bundle; and iii) stream delivery time (SDT) as the time
when the last packet of application data is delivered successfully at the receiver.
Note that SDT identifies the time when the last missing segment arrives at the
destination, which is also the time receipt of last missing DTN bundle.
The multimedia performance of BSS is evaluated based on the following three
metrics: i) end user’s display efficiency (EDE) is defined as the number of successfully
displayed video frames; ii) end user’s display attenuation as the number of wrongly
decoded video frames; and iii) average peak signal-to-noise ratio (PSNR) along with
actual snapshots of the video display.
116
Chapter 6: Evaluation Methodology
The results of the second simulation category, which includes all simulations that
are related to the proposed fully distributed and asynchronous data dissemination
model, are reported with the help of the following metrics: i) delivery ratio is defined
as the percentage of the created ADUs successfully delivered to all end-users (i.e.
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶∗𝐸𝐸𝐸𝐸_𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁
), ii) delivery latency as the average data latency between the
production time of ADUs on the satellite and their delivery to the end-users and iii)
data volume as the total amount of actual data, not including the headers, that was
delivered successfully to the end-users.
Finally, for the simulations conducted in the context of the third simulation
category, the following metrics are reported: i) guest-users performance is defined as
the aggregated average throughput achieved by the guest-users, ii) guest-users drops
as the amount of dropped guest-user data, iii) throughput impact as the impact of
guest-user traffic on home-user average throughput and iv) latency impact as the
impact of guest-user traffic on home-user traffic queueing delay. The impact values
are calculated according to the corresponding values attained by home users for the
same queue-management configuration, with zero guest users connected to the
access point.
6.5.
Experimentation Tools
6.5.1. Data Streaming Emulations
The emulation of the data streaming communication scenarios described in
subsection 6.3.1 was accomplished through the use of the ESA DTN testbed [131]
established in the Space Internetworking Center [132]. SPICE-ESA testbed provides
a realistic experimental environment for satellite and space communications,
including real and flight-ready components. Indeed, specialized hardware and
software components have been incorporated into the testbed, enabling the testing,
evaluation and validation of implemented mechanisms and protocols. Furthermore,
a link with a geostationary satellite, namely HellasSat 2, is utilized on demand, to
provide real satellite link characteristics for experimental purposes. The overall
architecture of the SPICE-ESA DTN testbed is presented in Figure 6-9.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 117
Figure 6-9. SPICE DTN testbed architecture.
In particular, for the purpose of the present evaluation study, the Linux kernel’s
embedded network emulator tool (Netem) and the one-way light time transmission
delay simulator tool (OWLTSIM), provided by ION, were used to emulate terrestrial
and Space networking conditions, respectively. Those tools are able to regulate
propagation delay and packet loss, and reliably emulate channel conditions.
118
Chapter 6: Evaluation Methodology
Equation (6-3) was used to calculate packet error rate values (PER) from the BER
values presented in subsection 6.3.1 and the bundle sizes employed in the
simulations. The calculated PER values were further adjusted to reflect the
difference in granularity between the two tools.
𝑃𝑃𝑃𝑃𝑃𝑃 = 1 − (1 − 𝐵𝐵𝐵𝐵𝐵𝐵)8∗𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 , bundle size in bytes
(6-3)
Equation (6-3) inherently assumes that independent bit errors and a Gaussian PER
distribution model accurately emulate channel noise. Given the wireless nature of
the communication links employed in the described scenarios, one might argue that
the Gilbert-Elliot error model is perhaps more appropriate because it also
accommodates burst errors, which are typical in wireless communication
environments. The counter-argument would be that in order to emulate bundle
layer noise reliably, one must keep in mind that burst error rates at the link layer do
not always translate into burst bundle losses. Moreover, the investigated
environments include transient disruptions that produce some of the same effects
as bursts of bit errors. In this context, the Gilbert-Elliot model is believed not to be
able to reflect in a satisfactory manner the nature of the environment in which the
experiments were conducted. However, this analysis may in the future require a
study on its own right.
Another issue of concern was the applicability of the PER computations in an
environment of asymmetric network links. When a bundle is forwarded through an
unreliable transport layer, such as UDP, it is encapsulated in its entirety within a
single datagram, while when the same bundle is forwarded through a reliable
transport protocol, it is fragmented into smaller entities, for example, TCP segments.
In order to evaluate the potential lack of accuracy that would result from applying a
uniform PER model to these two distinct cases, a simple worst-case analysis was
conducted [133] that showed that the loss in accuracy is minimal. Therefore, a
uniform PER model is applied for the emulations conducted in the context of this
experimental category.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 119
The emulation of the remaining elements, including network protocol stack
configurations and streaming applications, was based on the ION platform, version
3.1.1. As an implementation of the DTN architecture, fully conforming to RFC5050
[7] and RFC5326 [134], ION includes implementations of all the necessary
components used in the present study: BP, the Licklider Transmission Protocol
(LTP) [135] and both BSS framework and ipnfw.
Since both terrestrial and Space environments were emulated, two network stack
configurations were required. The ‘terrestrial’ stack employs Internet-based
network protocols—TCP/IP and UDP/IP—at the convergence layer. The
‘Space/deep Space’ stack utilizes LTP at the convergence layer: ‘green’ LTP
transmission functions as the ‘best-effort’ service while ‘red’ LTP transmission pro-
vides the ‘reliable’ service. In Figure 6-10, the complete stack of networking
protocols used in the experiments is depicted.
Figure 6-10. Representation of the network stack used in both configurations.
Two simple streaming applications were also developed in order to emulate the
behavior of actual streaming applications. These applications are designed to
acquire accurate timing measurements for the dispatch and reception of all ‘media’
bundles. bssStreamingApp simulates the functionality of a media source that
produces a stream of 30 frames per second. It wraps the produced stream in ‘media’
120
Chapter 6: Evaluation Methodology
bundles of user-defined size and hands them to the bundle layer for transmission.
Each bundle’s payload is a multiple of 1316 bytes in order to simulate the behavior
of the packetized elementary stream mechanism [136], which is used by popular
media streaming applications such as the VideoLan Client (VLC) media player [137].
At the other end, the bssRecv simulated media sink is based on the application
programming interface functions provided by the BSS library. It presents two basic
functionalities; it immediately displays any in-order ‘media’ bundles arriving at the
destination, and it saves, in a specially designed database, all ‘media’ bundles of the
received stream, whether received in-order or out-of-order; the bundles are sorted
by transmission order within the database. For the second set of experiments, since
BSS can be used as a transfer medium for multimedia streams, its preliminarily
multimedia performance also had to be tested. For that purpose, bssStreamingApp
and bssRecv were modified to act as middleware between the well-known media
player VLC and ION.
Throughout the whole experimental procedure, the transmission ratio achieved by
bssStreamingApp remained constant at 1 Mbps, which is considered adequate to
simulate the transmission of a H.264 [email protected] fps video quality stream. In
each case, it is assumed that for both terrestrial and Space communication links, the
available
bandwidth suffices
bssStreamingApp.
to
cover
the
transmission
rate
needs
of
Finally, BSSP applies an expiration timer to control the retransmission of lost
bundles. This timer is set equal to the round-trip time value of the communication
link plus a small safety margin in order to avoid any spurious timeouts resulting in
unnecessary retransmissions that could lead to the under-utilization of the available
bandwidth.
6.5.2. Asynchronous, LEO-based, Data Dissemination Model Simulations
In order to simulate the main scenario along with the associated variations and
corresponding
communication
patterns
described
in
subsection
6.3.2,
a
combination of STK [138], the physical simulator modeling of the satellite orbits,
Towards a Future Internet Architecture: Protocols and Support Mechanisms 121
and ns-2 [139], the network simulator modeling of the end-to-end network from the
satellite application to the end-users, was employed. The STK simulator calculates
the physical characteristics of the space contacts, including the start and end times,
bandwidth, bit error-rate and delay. Each contact includes the identification
numbers of the participating satellite and ground station, allowing association with
the corresponding entities in ns-2. The contacts are translated into ns-2 link up and
down events, which are then imported into ns-2 and incorporated in the end-to-end
topology.
The STK simulation takes into consideration the physical aspects of the satellite
communication scenarios, including the orbit geometry of the satellite, geographical
locations and coverage of ground stations, as well as the start epoch and duration
for each scenario. Link budgets for each contact between satellites and ground
stations are calculated using detailed models of the satellite transmitter and ground
station receiver, already available in STK. The STK simulation experiments are
based on the low-cost broadcast templates, which correspond to the space link
models described in section 4.3.
In ns-2 the entire topology is modeled as an IP network using UDP in space and TCP
on the ground. The central entity of the network simulation is an extended version
of the DTN agent developed by Dimitris Vardalis in ns-2 [140] and is based on the
DTN simulation model described in [141]. The DTN agent is deployed as an overlay
on the relevant nodes (i.e. satellites, ground stations, end-users, subscription
service), while the rest of the topology consists of pure IP nodes, relaying upper
layer traffic. The STK output is used in order to create links with the corresponding
characteristics between satellite and ground station nodes and also to set these
links to the up/down state accordingly during simulation. The simulator
implements five types of network nodes as described in section 4.3: Satellite,
Ground Stations, End Users, Subscription Service, and Uplink Ground Stations. All
these types are simulated by multiple instances of the DTN agent, appropriately
configured according to the desired simulated node type. The source DTN agent
exposes an interface that accepts the number of bytes to be transmitted and,
122
Chapter 6: Evaluation Methodology
optionally, the data ToS. Through this interface the agent creates ADUs, which are
then fragmented into one or more PDUs (i.e. bundles), based on the maximum
bundle size parameter. The bundles are inserted into a sending buffer of a user-
defined capacity. Newly arrived bundles may cause the eviction of old bundles in the
case of buffer space shortage.
6.5.2.1.
Data unit sizing
Ns-2 simulates network protocols at the packet level, hence computational
complexity is determined by the overall number of processed PDUs. Due to the high
level of simulation detail ns-2 is normally used for small-scale experiments. In order
to adapt it for large-scale simulations (i.e. length of 10 days and GBs of data), an
effort was channeled into lowering complexity while maintaining fidelity of the
results. In this spirit, experiments with larger PDU sizes were considered, reducing
the number of PDUs for a certain level of data production rate. However, increasing
the PDU size does not allow the use of the BER that is calculated by STK (i.e. for large
PDU size all PDUs would be corrupted). In order to overcome this problem the
Packet Error Rate (PER) that corresponds to the given BER was calculated for a
realistic nominal PDU size. The calculated PER was then applied to the larger PDUs
of the simulation yielding analogous error effect. The BER was converted to PER
using Equation (6-3).
Packet size, or bundle size as it appears in Equation (6-3), was set to the value of
65,000 bytes (approximately 64 KB) as this is the maximum size for a TM
encapsulation packet. The validity of the conversion approach was confirmed
through a series of short comparative simulations, using an always up, high-speed
link.
Based on the previous analysis the sizes of the ADU and PDU are set at 107 and 106
bytes (approximately 10 MB, 1 MB), respectively. In all cases an ADU consists of 10
PDUs.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 123
6.5.2.2.
Satellite operation
Data production is simulated by a custom application module that is configured to
produce an ADU of a certain size every some time period. The user sets the desired
daily data production rate as well as the ADU size and the simulation script
calculates the ADU production period. Each application tags the ADUs it creates with
a ToS number. An arbitrary number of application modules can be attached to the
same DTN agent, enabling the simulation of complex data generation patterns. ADUs
received at the DTN agent are fragmented into the appropriate number of PDUs and
entered into the buffer.
In the general case, satellites continuously broadcast to any ground station that may
be receiving. Satellite-deployed DTN agents use UDP as the underlying protocol of
the space link with ground stations. The packet/frame size overhead can be
configured based on the total overhead of the underlying protocols, independently
of the PDU header imposed by the higher layer protocol (i.e. CFDP, DTN, etc).
Data broadcasting is simulated through a group of point-to-point links from the
satellite to the ground stations. Each satellite node has links to all GS nodes. The
state of each link is set to up or down by utilizing the dynamic topology capabilities
of ns-2, according to the STK output. When setting a link to the up or down state the
corresponding DTN agent virtual connections are created or destroyed, respectively.
The point-to-point approach was preferred over a pure broadcast method (i.e. using
a wireless protocol such as 802.11), since it allows for precise control over the space
link characteristics (bandwidth, delay, error rate).
6.5.2.3.
Transmission scheduling
Data received by the satellite are packaged into Application Data Units (ADUs) of a
specified size (see subsection 6.5.2.1) and passed to the broadcasting subsystem. In
order for an ADU to be created the appropriate amount of data must be first
acquired. The size of the ADU must be greater than or equal to the minimum useful
granule of data for a certain application. For performance reasons, multiple data
units may be combined into a single ADU, especially in case of extremely small data
124
Chapter 6: Evaluation Methodology
units (e.g. emergency signals). The broadcasting subsystem fragments incoming
ADUs into a number of PDUs, according to a specified PDU size, and inserts these
PDUs into the transmission buffer. Old PDUs may need to be removed in case a
certain buffer threshold is exceeded, thus imposing an implicit data TTL.
Consequently, data TTL plays an important role in this process as it dictates how
retransmission effort will be distributed among old and recently created PDUs. A
high TTL value would favor older PDUs while a low TTL value would favor newer
PDUs.
The transport agent at the satellite continuously broadcasts PDUs at a rate imposed
by the available bandwidth. Newly created data are given higher priority over data
that have been transmitted at least once, so new PDUs are transmitted as they arrive
in a first-come-first-served order. When all data have been transmitted at least once
an ADU is randomly selected from the buffer and transmission resumes with the
first PDU belonging to the selected ADU. Transmission continues sequentially until
all PDUs belonging to the selected ADU have been transmitted, at which point a new
ADU is randomly selected for retransmission and the process iterates. If at any point
during the randomized retransmission phase new PDUs arrive, the process is
interrupted and the new PDUs are transmitted first. Figure 6-11 shows an example
case where the transmission buffer contains 4 ADUs of 5 PDUs each. While
transmitting PDU 3 of ADU 2, a new ADU arrives (ADU 4) interrupting the
retransmission of ADU 2 so that ADU 4 can be transmitted for the first time. As soon
as all PDUs belonging to ADU 4 have been transmitted, retransmission of ADU 2
resumes with the transmission of PDU 4. When ADU 2 finishes retransmitting (i.e.
PDU 5 is transmitted), one of the four stored ADUs is randomly selected for
retransmission and the process repeats. If reception feedback is not available, as in
the scenario described above, it is possible that PDUs already received on the
ground will be retransmitted, potentially wasting downlink bandwidth.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 125
Figure 6-11. Transmission buffer example with 4 ADUs of 5 PDUs each.
6.5.3. Broadband Sharing Simulations
The simulation of the broadband sharing main scenario along with its associated
variations described in subsection 6.3.3 was accomplished by employing a number
of data collection, analysis and simulation tools. With respect to the PAWS project
trial data and their corresponding analysis presented in section 6.2, the following
tools were employed: Netperf [142] was used to measure the upload and download
throughput (Mb/s) values of the access points employed in the PAWS project trial.
The end-to-end RTT values from those access points to a set of servers in the UK
were collected using the ping tool. The Broadband Internet Service Benchmark
(BISmark) tool [143] was used to obtain the respective lmrtt, ulrttdw and ulrttup
values of each access point, while tcpdump [144] was employed to account use of
PAWS service. The respective flow-level characteristic samples from the collected
tcpdump files were extracted by using the tcptrace tool [145].
The simulation procedure per se was based on the ns-2 simulator, where the
topology presented in Figure 6-8 was modeled as an IP network. Each one of the
presented network elements (i.e. home users, guest users, ADSL access point, the
Internet, servers) was represented as a single network node, with corresponding
network properties, as indicated by the analysis results presented in subsection
6.2.2.3 and the respective scenario assumptions. All home and guest users employ
standard TCP agents for transferring data to the server elements. Specifically for
guest users, a new network application was developed in order to accommodate the
production of network traffic with specific statistical properties. Essentially, this
new network application is a versatile traffic generator that is able to generate TCP
126
Chapter 6: Evaluation Methodology
flows of specific size and duration over different time intervals based on various
statistical distributions. Any NS2 TCP agent can be attached to this traffic generator,
while the statistical distributions for the experiments conducted in the context of
the broadband sharing simulation category regarding each one of the
aforementioned properties are specified in subsection 6.2.2.3.
7. Evaluation Results
7.1.
Experimental Category I – Data Streaming
7.1.1. Network Performance
In this first phase of BSS’s evaluation, its network performance was measured by
employing the SDE, SDA and SDT metrics. SDE and SDA results for all three
terrestrial scenario subcases are presented in Figure 7-1. The interval at which the
SDE and SDA metrics were monitored was set to 5 minutes, an adequate time period
considering the relatively low propagation delay values of the corresponding
terrestrial network topologies. The source continuously transmits “media” bundles
to the destination for an indefinite period of time.
As the results presented in Figure 7-1 suggest, BSSP outperforms IPN forwarder in
the majority of the cases, managing to deliver a larger number of bundles within the
specified timeframe. In typical terrestrial networking environments, represented by
the Normal subcase, both forwarding approaches achieve equal performance,
successfully delivering the expected number of bundles to the final destination.
However, in highly-stressed terrestrial networking environments, represented by
the Low and High subcases, the SDE values indicate that the BSSP minimizes TCP
performance degradation and gains a noticeable advantage over the IPN forwarder.
127
128 Chapter 7: Evaluation Results
(a)
(b)
Figure 7-1. Comparison between BSSP and IPN forwarder for the a) 3-node and b) 5-node
network topology of all three terrestrial scenario variations.
The comparative impact on streaming performance becomes even more significant
as the payload size increases. The explanation is straightforward: smaller packets
are less susceptible to channel noise than larger packets, so in the large-payload
case, where more bundles are getting lost, IPN forwarder expends more time for
recovery. Unlike IPN forwarder, BSSP manages to deliver higher volumes of bundles,
since it avoids all unnecessary retransmissions at this stage. However, it should be
noted here that BSS’s SDE improvements come at some cost. IPN forwarder
naturally achieves 100% in-order delivery of bundles, since it relies solely on TCP
mechanisms; there is no need ever for a bundle to be re-forwarded. The BSSP must
re-forward all bundles that were originally forwarded via UDP but were lost in
transit. The net effect is that the BSSP does less TCP transmission than IPN
forwarder but must do additional bundle forwarding, ranging from 6% up to 40%
more. This outcome is really not as dramatic as it may initially appear to be: the
Towards a Future Internet Architecture: Protocols and Support Mechanisms 129
overall performance impact of this additional transmission is low considering the
performance cost savings of initial UDP forwarding compared to TCP transmission.
The most important point on which one should focus here is the vast improvement
in the total number of delivered bundles.
The rest of the SDE and SDA quality metrics results are produced by the three Space
scenarios and presented in Figure 7-2. Although the comparative evaluation results
are explained by similar arguments, Space scenarios incorporate two distinct
variations from the terrestrial scenarios. The long-haul propagation delay modeled
in most Space scenarios suggested an increased sampling interval (i.e. 10 minutes)
for SDE and SDA metrics. This was deemed necessary in order to allow for sufficient
time for the participating nodes to complete at least one communication round.
Furthermore, Space links are typically associated with low data rates, such that
larger bundle payloads are preferable in order to avoid unnecessary overhead.
Therefore, the examined payload sizes were 2632 bytes and 3948 bytes.
From the experimental results presented in Figure 7-2, it is clear that BSSP achieves
equal or better performance than IPN forwarder in all cases and for both variants of
the topology. In particular, an improvement ranging from 20% up to 100% is
observed in SDE values for both Spacecraft and Mars scenarios. Beyond the
improvement in the delivery efficiency of bundles, BSSP manages also to increase
the quality of the received stream by significantly decreasing the value of SDA in all
cases. The difference in magnitude of the SDA values between the Spacecraft and
Mars scenarios exists due to the asymmetric topology of the Mars scenario, which
incorporates a long-haul propagation link at the first hop of transmission and
results in fewer bundles reaching the intermediate node, thus decreasing the total
number of retransmissions. This results in fewer out-of-order packets received at
the final destination.
However, fewer bundles than expected were delivered in total in the Spacecraft and
Mars tests using 2632-bytes payloads. This could be attributed to the heavy
computational load imposed by LTP’s retransmission accounting in these tests:
130 Chapter 7: Evaluation Results
because the bandwidth delay products are large, a large number of small
transmission segments must be tracked.
(a)
(b)
Figure 7-2. Comparison between BSSP and IPN forwarder for the a) 3-node and b) 5-node
network topology of all three Space scenarios.
Based on the outcome of the quality metrics, SDT metric was additionally employed
to further assess the efficacy of the BSSP regarding delivery time. For this purpose,
SDT was measured over a stream consisting of 5000 “media” bundles. The
streaming session was deemed completed when all 5000 bundles were successfully
received at the final node. The purpose of these experiments was to map in time the
change in SDE values observed during the first set of experiments.
BSSP superiority was further confirmed by this second set of results (Figure. 7-3).
Even after incorporating the additional error margin into the results, BSSP still
outperforms ipnfw across the entire set of experiments. The reduction in the total
requested time for successfully transferring 5000 frames at the destination node is
Towards a Future Internet Architecture: Protocols and Support Mechanisms 131
significant for both terrestrial and Space scenarios, with a performance gain ranging
from 20% up to 130%. Finally, based on the comparison of the results acquired by
the 3-node and 5-node topology variants respectively, it was confirmed that the
addition of extra nodes in isometric distances along the end-to-end path of a longhaul communication link with stable BER has a positive effect on the performance of
both forwarding approaches.
(a)
(b)
Figure 7-3. Comparison between BSSP and IPN forwarder based on SDT metric of a
representative sample of cases from both a) terrestrial and b) Space scenarios.
7.1.2. Multimedia Performance
For the multimedia performance evaluation, the well-known VLC media software
was used as the basic media platform for broadcasting and receiving an actual 2-
minute video clip.
(a)
(b)
(c)
Figure 7-4. Comparison between BSSP and IPN forwarder based on a) EDE b) EDA and c)
PSNR metrics for the 3-node network topology of the Moon Space scenario.
Based on the results presented in Figure 7-4, it becomes apparent that BSS achieves
a notable multimedia performance improvement by delivering a higher number of
132 Chapter 7: Evaluation Results
media bundles in the correct order. Practically, this leads to the successful display of
almost 18% more video frames, based on the comparison of EDE values, and at the
same time to a significant decrease in the number of wrongly-decoded frames. In
general, PSNR values below 30dB suggest very poor image quality. This fact can be
attributed to the high BER of the communication channel. Nevertheless, applying
BSS resulted in the increase of PSNR value from 12dB to 14dB, a fact that further
indicates an improvement in the real-time viewing experience in comparison to that
offered by IPN forwarder.
An indicative sample of the actual real-time display images under comparison is
presented in Figure 7-5. Although both images are severely distorted, the quality of
the first image enables the viewer to clearly identify the object in focus. Such a
difference in image clarity could play a key role in real use cases such as the
accurate assessment of a potential emergency situation from the end-user of a
monitoring system.
(a)
(b)
Figure 7-5. Video stream snapshots of a) BSSP and b) IPN forwarder for the 3-node
network topology of the Moon Space scenario.
7.2.
Experimental Category II – Asynchronous, LEO-based, Data
Dissemination Model
Figure 7-6 presents the results of the first test case, which clearly show that
employing more ground stations has a positive impact on the overall performance of
the system. Delivery latency is significantly decreased while all other key
performance indicators, including delivery ratio, are improved. Furthermore, a
trade-off between the number of the employed ground stations, closely associated
Towards a Future Internet Architecture: Protocols and Support Mechanisms 133
to their respective geographical locations, and the improvement of system’s
performance is also revealed. From a critical point (i.e. 100 ground station units in
this particular case) and onwards, the increase in the number of ground stations
Delviery Ratio (%)
impact on the overall system performance.
Delivery Ratio
100
80
4.5
4
3.5
60
3
40
20
0
2.5
52
100
140
Ground Station (units)
2
1.5
Delivery Latency (h)
providing downlink services to the LEO satellite does not seem to have a significant
Delivery Ratio (%)
ground stations.
Delivery Ratio
100
80
600
500
400
60
300
40
200
20
0
1
6
Satellite (units)
46
100
Data Volume (GB)
Figure 7-6. Low-cost broadcast delivery ratio and delivery latency varying the number of
0
Figure 7-7. Low-cost broadcast delivery ratio and data volume varying the number of
satellites.
The results regarding the second test case are presented in Figure 7-7. One can
observe that the data return increases linearly with the number of employed
satellites for the 1 and 6 satellite cases (i.e. when 6 satellites are employed the data
return is 6 times higher than the single satellite configuration). This linear increase
does not hold true for the 46 satellites configuration where data return of the
system improves only by a factor of 3.8 vs. a 7.7 increase in the number of satellites
compared to the 6 satellites configuration. This fact indicates that ground stations
134 Chapter 7: Evaluation Results
network is transformed into a bottleneck; an argument that it is also supported by
the low values of delivery ratio metric in comparison to the previous cases (i.e.
single and 6 satellites cases). Since the creation of the bottleneck cannot be related
to the ground station link bandwidth, the total number and duration of contact
opportunities between the satellites and ground stations seems to be the only
remaining factor that could cause this issue. This indicates the existence of a tradeoff between the number of employed satellites and the number/location of the
ground stations providing downlink service.
7.3.
Experimental Category III – Broadband Sharing
Figure 7-8 outlines the simulation results for the third experimental category. Each
row of panels presents the results for access network profiles AN1, AN3, AN8 and
AN16, respectively. More specifically, the results concerning the single guest-user
case show that the impact of guest-user traffic on home-user network performance
remains relatively low across all queue-management configurations and for all
access networks. In particular, the largest impact values observed on home-user
traffic average throughput and queueing delay are 4% and 4.5ms, respectively. This
fact can be attributed to the low-bandwidth demands of guest users, as shown in
subsection 6.2.2.3. The CBQ configuration in AN16 (see Figure 7-8(h)) constitutes
the sole exception to these observations, as the impact of guest-user traffic on home-
user traffic queueing delay reaches up to 7.5ms. This clearly indicates that a
potential decision to exclusively allocate resources for guest-user traffic in systems
with limited capacity might cause considerable impact on home user’s network
performance, even when the actual guest-user traffic is low.
The same set of results pertaining to the single guest-user case also suggests that
there is a considerable deviation in the service ratios of guest-user traffic among the
various investigated queue-management configurations. In particular, for access
network profiles AN1 and AN3, all queue-management configurations, except from
UPNQ configuration, serve guest-user traffic adequately, while causing almost zero
guest-user data drops. Due to the heavy network traffic generated by home users,
Towards a Future Internet Architecture: Protocols and Support Mechanisms 135
UPNQ configuration follows a strict approach by allocating zero bandwidth on
serving guest-user traffic. For access network profiles AN8 and AN16, in which
upload resources are scarce, HPSS and PRIO configurations resemble UPNQ, by
following the same strict approach. The rest queue-management configurations
derive similar behavior as in access network profiles AN1 and AN3.
In contrast to the single guest-user case, system behavior among the different
queue-management configurations differs significantly in the multiple guest-user
cases. The first conclusion that can be drawn is that, in the specific context of
domestic broadband sharing schemes, the DropTail configuration constitutes an
inappropriate queue-management choice. In particular, DropTail configuration
allows excessive amounts of guest-user traffic to be served – even when the network
is congested - causing considerable impact on home users’ network performance.
The impact grows further as the number of guest users increases or/and the uplink
bandwidth decreases, reaching up to 57%.
Interestingly, similar behavior is observed for three other queue-management
configurations, which are based, respectively, on RED, CoDeL and SRR algorithms.
This behavior can be attributed to the fact that none of these queueing disciplines
are capable of differentiating traffic based on traffic classes. Therefore, they
randomly penalize flows belonging to both home- and guest-user traffic in their
effort to either enforce system fairness or preserve low queue latencies. Due to
TCP’s congestion-avoidance and control mechanisms, penalizing a home-user
elephant flow forces the sender to significantly reduce its window size. In the
scenario considered in the context of the present evaluation study, such an action
causes a severe degradation of the overall home users’ network performance.
Contrary to home-user traffic, penalizing a specific guest-user flow would not have
the same devastating impact in the total volume of traffic originating from guest
users, given that most guest-user packets belong to mice flows. Thus, independently
of the penalty that either AQM- or FQ-based queue disciplines impose on guest-user
flows, guest-user traffic will end up consuming a significant portion of system
resources, as it is also shown in panels (a), (c), (e) and (g) of Figure 7-8.
136 Chapter 7: Evaluation Results
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
Figure 7-8. Broadband sharing simulation results. Panels (a), (c), (e) and (g) present guest-
user aggregated average throughput on the primary y-axis, with the respective impact
Towards a Future Internet Architecture: Protocols and Support Mechanisms 137
values on home-user average throughput presented on the secondary y-axis. Panels (b), (d),
(f) and (h) present the amount of guest-user dropped data on the primary y-axis, with the
respective guest-user traffic impact on home-user traffic queueing delay presented on the
secondary y-axis. To secure the distinctiveness of the figures, the results regarding CoDel
have been omitted, due to their high similarity with the behavior presented by RED.
Unlike Droptail, AQM- and FQ-based approaches, priority- and class-based-queueing
approaches seem to effectively mitigate the contradictive network requirements
between home and guest users. They achieve that by serving a certain amount of
guest-user traffic, without imposing large impact on home-user network
performance. Among PRIO, UPNQ, HPSS and CBQ configurations, UPNQ
configuration constitutes the strictest method, since it allocates the least amount of
bandwidth for serving guest-user traffic. PRIO is the second strictest method. HPSS
configuration presents a good overall performance across most cases. It achieves a
good balance between serving guest-user traffic and degrading home users’ network
performance. In particular, the imposed impact on home users’ network
performance is low, ranging from 0 up to 1.8%, regarding the aggregated average
throughput, and less than 2.5ms, regarding queueing delay. CBQ configuration could
also be considered adequate choice, given that wider queueing delay tolerance
levels are acceptable. In particular, CBQ configuration allocates the largest amount
of network resources for serving guest-user traffic in comparison with all other
priority- or class-based-queueing configurations. At the same time though, it
imposes a moderate queueing delay impact on home users’ network performance,
which in our test cases, reaches up to 12ms.
In terms of system scalability, the results indicate that domestic broadband sharing
networks that are supported by high-bandwidth connectivity links could serve
simultaneously up to 5 guest users even under high-congestion conditions. These
types of domestic broadband sharing networks are represented in the present study
by network access profiles AN1 and AN3. The average throughput provided per
guest user in those networks, considering that either HPSS or CBQ configuration is
employed, is 1.2KBps. This value is within the average throughput values range
138 Chapter 7: Evaluation Results
reported in subsection 6.2.2.3; therefore, it can be safely argued that the notion of
service for guest users is preserved in these cases. However, the same is not valid for
the broadband sharing networks supported by low-bandwidth connectivity links.
The average throughput achieved in those networks, considering again that a
queue-management configuration, which does not impose large impact on home
users is employed, is just a few bytes per second. At this point, it should be noted
that these values correspond to a worst-case-scenario situation. It is reasonable to
expect that in actual deployments such worst-case-scenario conditions will occur
rarely or for short timeframes, allowing guest users to generally enjoy a decent
network experience.
Another aspect that should be carefully considered before selecting a queue-
management configuration for home broadband sharing schemes is its overall
dropping behavior with guest-user traffic. Recent trends [146] suggest that
smartphones and tablets will drive an even greater portion of future Internet traffic.
This percentage is expected to be greater among the users of open Internet access
services, due to their opportunistic connectivity nature and the high availability of
mobile devices. Queue-management configurations that cause large amounts of
guest-user data to be dropped might affect the battery life of guest users’ mobile
devices negatively due to the excessive retransmission effort undertaken by clients
to recover data losses. The results presented in Figure 7-8 also reveal a trend in this
topic, specifically regarding class-based-queueing approaches. This trend suggests
that the more resources allocated by this type of queueing approaches for serving
guest-user traffic the larger the amount of data drops for guest users, given that
guest-user resource demands are high. In particular, as more resources are
allocated for serving guest-user traffic, guest-user queue will take more time to be
filled up. In this time period, guest users will have further increased their
transmission windows. Larger transmission windows result in more catastrophic
data-loss events, once congestive collapse occurs.
Finally, two additional sets of simulations were performed to verify the robustness
of the acquired simulation results, under different parameter values for the
Towards a Future Internet Architecture: Protocols and Support Mechanisms 139
statistical distributions that were used to model guest-user uplink traffic. For the
first set, the parameter values presented in Table 6-1 were decreased by 10%,
whereas for the second set, they were increased by 10%. All other simulation
parameters were left intact. The new results confirm that both the qualitative
characteristics of the investigated queue-management configurations and all the
associated observations regarding the original set of experiments remain valid for
both cases. Of course, the actual values of most metrics present some differences in
comparison with the original set, mainly due to the associated changes in the
volume of guest-user traffic.
140 Chapter 7: Evaluation Results
8. Conclusions and Open
Issues
The main focus of this thesis has been laid on the development of novel solutions
regarding a number of challenging issues related to the Future Internet. These issues
were selected on the basis of existing limitations and future expectations: the
current Internet architecture is expected to evolve rapidly in the following years by
interconnecting different types of devices operating in heterogeneous network
environments, supporting myriads of applications and providing ubiquitous
network connectivity, independently of location and user privileges. With this in
mind, special emphasis was given on data dissemination, availability, and
accessibility topics.
In this context, Delay-/Disruption-Tolerant Networking was proposed as a network
architecture that is suitable to support the formation of the Future Internet. More
specifically, it was argued that DTN has the potential to play a key role in this
formation process: first, by allowing users and devices from heterogeneous
networking environments to transparently communicate with each other,
independently of the underlying protocol stack; and, second, by acting as a service
layer for mitigating various network connectivity issues, such as long delays and
connectivity disruptions, that occur mainly due to the continually extended use of
wireless communications.
141
142 Chapter 8: Conclusions and Open Issues
With respect to the development effort conducted in the course of this thesis, it was
mainly invested towards network infrastructure solutions, while only subparts of
the developed solutions were dedicated to the end-users. Overall, this effort
resulted in the following outcomes:
i)
Enhancing Future Internet data dissemination capabilities through data
streaming content distribution mechanisms: In particular, it was shown
that it is possible to efficiently transfer data streams over heterogeneous
internetworked environments. To this end, Bundle Streaming Service was
developed as a proof of concept, demonstrating that a universal, DTN-
based, data-streaming framework could be employed to alleviate most of
the networking challenges related to both live and stored data streaming
ii)
over heterogeneous network environments.
Broadening Future Internet availability via an asynchronous, satellitebased data dissemination model: Through the design of a fully-
distributed, scalable, LEO-based, data dissemination system, it was shown
that it is possible for Space assets to be seamlessly incorporated into a
Future Internet architecture, in order to broaden its availability and
widen the accessibility of its resources.
Apart from the integration of Space assets to a unified Future Internet
architecture, a fact that certainly helps boost network availability and
accessibility, another aspect of the proposed system that further
enhances both of these properties is its ability to asynchronously deliver
data to interested end-users across the globe. This was achieved with the
incorporation of several DTN characteristics into the proposed system in
conjunction with the extension of traditional satellite communications in
order to accommodate low-cost, dense satellite networks with limited or
no uplink opportunities. As a result, this particular approach significantly
increases the cost-effectiveness of the proposed system, thus, making it
affordable and applicable to a much wider audience.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 143
iii)
Widening the accessibility on Future Internet network resources by
taking advantage of resource pooling techniques. More specifically, it was
shown that it is possible to exploit unused bandwidth capacity from
current network infrastructures in order to strengthen the overall global
networking participation and enhance the social impact of the Future
Internet. In particular, the research effort focused on broadband sharing,
a common resource pooling technique with high scalability potential that
can be applied to almost every available access point. In this context, after
numerically assessing the capability of broadband sharing systems in
providing an adequate level of service to guest users, Hybrid Packet
Scheduling Scheme was introduced, as a queue-management framework
that allows guest users to exploit the available network resources more
effectively. HPSS enables Internet access to be further broadened in a
transparent manner to the end-users without any additional cost.
Extensive analytical, simulation and emulation studies were performed, accordingly,
to assess both the efficiency and the effectiveness of the aforementioned solutions.
The respective evaluation results provided useful insights into the dynamics of the
various Future Internet topics explored herein. One such result, which is applicable
in broader scope beyond the context boundaries of this thesis, refers to the
scalability of broadband sharing systems in terms of the number of supported guest
users. In particular, the evaluation results indicated that high-bandwidth domestic
broadband sharing networks can serve simultaneously a considerable number of
guest users even under high-congestion conditions. This scientific conclusion clearly
broadens the perspective of broadband sharing dynamics and prescribes
appropriate tools to be employed as the medium for offering free Internet access.
The rest of this chapter expands on the detailed conclusions drawn from the results
for each one of the proposed solutions.
144 Chapter 8: Conclusions and Open Issues
8.1.
Bundle Streaming Service
The potential contribution of the BSS framework to enhance data streaming
capability among highly heterogeneous network environments was assessed based
on extensive experiments. Both networking and multimedia aspects of BSS were
evaluated. A series of results were obtained that demonstrate a clear-cut advantage
for BSS in a number of cases. The range of improvement observed depends on the
particular network conditions. More specifically :
1) The duality of BSSP, with both reliable and unreliable transmission
capability, allows for a smoother near real-time streaming viewing
experience, even in highly disrupted communication environments.
2) The BSS framework improves stream reception in both terrestrial and
Space environments.
3) The BSS framework is clearly advantageous for long-haul communications.
4) The BSS framework’s API facilitates the development of data streaming
applications in heterogeneous network environments.
Armed with the knowledge acquired by these evaluations, the majority of future
plans include actions to further enhance BSS capabilities and to confirm the validity
of emulation results in real wireless environments. Examples of these actions
constitute the investigation of BSS’s adaptability to multicasting communications, its
deployment in embedded devices and its evaluation in real Space environments.
8.2.
Asynchronous, LEO-based, Fully-distributed Data Dissemination
System
With respect to the proposed data dissemination system, interesting conclusions
can be drawn from both system design and performance evaluation perspectives.
Regarding the introduced system design, it demonstrates in a concrete manner that
the incorporation of an asynchronous, fully-distributed, Space-based data
dissemination model into a broader Internet architecture is technically feasible by
utilizing
already-existing
and
proven-to-work
networking
technologies.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 145
Furthermore, it showcases DTN architecture capabilities in extending network
systems’ functionalities, while, at the same time, reducing systems’ complexity; thus,
providing a decisive contribution to the systems’ overall efficiency.
From a system performance point of view, simulation experiments showed that
when multiple, low-cost satellites are employed the performance gap related to the
total data return between the proposed system and traditional, high-performance,
point-to-point, LEO satellite approaches closes, rendering the deployment of a low-
cost broadcasting system a promising alternative in economic terms. In particular,
for the 46-satellite case the total volume of data delivered on the ground reached
562GB. This value is approximately 1/7 of the maximum 3.9TB amount of data
usually delivered by traditional, high-performance, point-to-point LEO satellite
systems, as indicated in [9]. Apparently, a direct comparison in terms of
performance between the proposed vs. the traditional approach would clearly favor
the latter. This might not necessarily be true though, in case economic data were
additionally utilized for the calculation of the cost-per-byte in each approach. Such
analysis is not part of the present thesis, but it may present an interesting possibility
for the continuation of this work. By and large, the results presented in this thesis
are, indeed, promising, considering the vast gap between the specifications of the
two systems. In any case though, further consideration and evaluation of the
suggested system is required to advance the planning and implementation strategy
and to study the extent to which its performance can be improved.
8.3.
Hybrid Packet Scheduling Scheme
In the context of employing broadband sharing as a medium for widening the
accessibility on Future Internet network resources, the behavior of queueing
systems administering the resource sharing process of an access point between
home and guest users was assessed from both an analytical and an experimental
perspective.
The numerical analysis was mainly employed as a tool for highlighting the role of
certain broadband sharing model aspects, such as the packet-size distribution of the
146 Chapter 8: Conclusions and Open Issues
traffic flowing through the access point and link speeds supporting access point’s
operation. In particular, it was shown that for low link speeds the application of a
priority-queueing system alone is not sufficient; thus, an additional mechanism for
regulating the arrival rate of the low-priority traffic is required to guarantee non-
significant impact on high-priority traffic. For high link speeds, a traffic pattern that
permits a certain amount of access point bandwidth to be exclusively allocated for
serving guest-user traffic was analytically identified. To effectively support both
low-speed and high-speed link configurations on broadband sharing systems,
Hybrid Packet-Scheduling Scheme was introduced.
Based on the aforementioned observations, an experimental assessment was
conducted. The aim was to evaluate the functional capabilities of a wide set of
queue-management configurations in distributing domestic broadband sharing
network resources between home and guest users. The corresponding results clearly
demonstrated the inability of AQM- and FQ-based approaches to mitigate effectively
the occasionally contradictive network requirements between home and guest users.
Instead, it was shown that priority- and class-based-queueing approaches could
better serve this purpose, by enabling guest users to access system resources in a
less-than-best-effort manner. Among these approaches, HPSS exhibited the most
balanced behavior by serving a considerable portion of guest-user traffic without
degrading home users’ network performance. In terms of system scalability, it was
shown that in certain cases, it is possible to serve up to five guest users
simultaneously, even under highly congested network conditions. Finally, apart
from the various performance aspects, a trend was identified, which suggests that
inappropriate configuration of class-based-queueing approaches could lead to
severe data losses for guest users and degrade their quality of experience by
affecting the battery life of their mobile devices negatively.
Ultimately, the conducted work complements previous analytical and experimental
studies on this topic [116] [117], and paves the way for the development of more
sophisticated
queue-management
configurations
for
domestic
broadband
connection sharing schemes. In future work, the plan is to expand the results
Towards a Future Internet Architecture: Protocols and Support Mechanisms 147
presented in this thesis with richer guest-user traffic datasets [147] and investigate
the potential use of delay tolerant networking concepts for handling guest-user
traffic at gateways, under periods of heavy network usage from home users. In
particular, the aim is to consider delay tolerant queue-management architectures,
similar to the one presented in [148], for caching guest-user content at gateways
and time-shifting its service in time periods where home-user traffic is low.
148 Chapter 8: Conclusions and Open Issues
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
B. Sanou, “ICT Facts and Figures,” ICT Data and Statistics Division,
International Telecommunication Union (ITU), Geneva, May 2015.
D. Evans, “The Internet of Things-How the Next Evolution of the Internet Is
Changing Everything,” White Paper, Cisco Internet Business Solutions Group
(IBSG), April 2011.
S. Burleigh, V. Cerf, J. Crowcroft, and V. Tsaoussidis, “Space for Internet and
Internet for Space,” Ad Hoc Networks, Elsevier, Special Issue on New Research
Challenges in Mobile, Opportunistic & Delay-Tolerant Networks, pp. 80–86,
2014.
L. Wood. Satellite Constellation Networks, “Chapter 2: Satellite Constellation
Networks,” Internetworking and Computing over Satellite Networks Book,
Edited by Y. Zhang, pp 13-34, March 2003.
P. Gill et al., “Characterizing Web Censorship Worldwide: Another Look at the
OpenNet Initiative Data,” ACM Transactions on the Web (TWEB), ACM, Vol. 9,
no. 1, pp. 1-29, January 2015.
V. Cerf et al., “Delay-Tolerant Networking Architecture,” RFC 4838, April 2007.
K. Scott and S. Burleigh, “Bundle Protocol Specification,” RFC 5050, November
2007.
S.-A. Lenas, S. C. Burleigh, and V. Tsaoussidis, “Bundle Streaming Service:
Design, Implementation and Performance Evaluation,” Transactions on
Emerging Telecommunications Technologies, Vol. 26, Issue 5, pp.905-917, May
2015.
D. Vardalis, S.-A. Lenas, S. Diamantopoulos, V. Tsaoussidis, F. Belli, P. Beavis, D.
Patterson, and K. Nergaard, “Decentralized Space-Data Dissemination for LowCost, Dense Satellite Networks,” IEEE Transactions on Aerospace and Electronic
Systems, 2015 (to appear).
[10] S.-A. Lenas and V. Tsaoussidis, “Enabling free internet access at the edges of
broadband connections: a hybrid packet scheduling approach,” SIGMOBILE
Mobile Computing and Communications Review (MC2R), Vol. 18, Issue 1, pp.5563, Feb. 2014.
[11] V Cerf, “The Internet is for Everyone,” RFC: 3271, April 2002.
[12] S. Schütz, L. Eggert, S. Schmid, and M. Brunner, “Protocol enhancements for
149
150 References
intermittently connected hosts,” SIGCOMM Computer Communication Review,
Vol. 35, Issue 3, pp. 5-18, 2005.
[13] J. Ott and D. Kutscher, “A Mobile Access Gateway for Managing Intermittent
Connectivity,” In Proceedings of 14th IST Mobile & Wireless Communication
Summit, Dresden, June 2005.
[14] D. Murray and T. Koziniec, “The state of enterprise network traffic in 2012,” In
Proceedings of 18th Asia-Pacific Conference on Communications (APCC), IEEE,
pp. 179-184, 2012.
[15] J. Choi, C. Park, H. Jung, T. Kwon, and Y. Choi, “Addressing in Future Internet:
Problems, Issues, and Approaches,” In Proceedings of 3rd International
Conference on Future Internet Technologies, June 2008.
[16] L. Zhang, R. Wakikawa, and Z. Zhu, “Support mobility in the global internet,” In
Proceedings of the 1st ACM workshop on Mobile internet through cellular
networks (MICNET '09), New York, NY, USA, pp. 1-6, 2009.
[17] Ram Kumar Ravikumar Santhi, “A Service Based Approach for Future Internet
Architectures,” Doctoral dissertation, UNIVERSITY OF AGDER, Faculty of
Engineering and Science, 2010.
[18] EC FIArch Group, “Fundamental Limitations of current Internet and the path to
Future Internet,” March 2011.
[19] R. Hinden, S. Deering, “IP Version 6 Addressing Architecture,” RFC4291,
February 2006.
[20] C. Perkins, “IP Mobility Support for IPv4, Revised,” RFC5944, November 2010.
[21] R. Moskowitz, and P. Nikander, “Host Identity Protocol (HIP) Architecture,”
RFC4423, May 2006.
[22] S. Bradner, L. Conroy, and K. Fujiwara, “The E.164 to Uniform Resource
Identifiers (URI) Dynamic Delegation Discovery System (DDDS) Application
(ENUM),” RFC6116, March 2011.
[23] S. Paul, J. Pan, and R. Jain, “Architectures for the future networks and the next
generation Internet: A survey,” Computer Communications, vol 34, no. 1 pp. 242, 2011.
[24] D. Fisher, “A look behind the future internet architectures efforts,” SIGCOMM
Computer. Communications Review, vol 44, no 3, pp. 45-49, 2014.
[25] J. Pan, S. Paul, R. Jain, “A survey of the research on future internet
architectures,” Communications Magazine, IEEE , vol.49, no.7, pp.26,36, July
2011.
[26] T. Zahariadis et al., “Towards a Future Internet Architecture,” In Proceedings of
Domingue, J., et al. (eds.) FIA 2011. LNCS, vol. 6656, pp. 7–18, Springer,
Heidelberg, 2011.
[27] L. Peterson (Ed) et.al, “GENI design principles,” Tech. Rep. 06-08, GENI Design
Towards a Future Internet Architecture: Protocols and Support Mechanisms 151
Document, August 11, 2006.
[28] A. Gavras, A. Karila, S. Fdida, M. May, and M. Potts, “Future internet research
and experimentation: the FIRE initiative,” SIGCOMM Computer
Communications. Review, vol. 37, no 3, pp. 89-92, 2007.
[29] J Wyatt et al., “Disruption Tolerant Networking Flight Validation Experiment
on NASA's EPOXI Mission,” In Proceedings of the First International Conference
on Advances in Satellite and Space Communications (SPACOMM 2009), pp. 187
– 196, Colmar, France, 2009.
[30] S. Parikh, and R. C. Durst, “Disruption tolerant networking for Marine Corps
CONDOR,” In Proceedings of Military Communications Conference (MILCOM
2005), IEEE, Atlantic City, NJ, USA, pp. 325 – 330, 2005.
[31] S. Saha, A. Sheldekar, A. Mukherjee, S. Nandi, “Post disaster management using
delay tolerant network,” In Proceedings Of Recent Trends in Wireless and Mobile
Networks, Springer, pp. 170–184, 2011.
[32] E. Magistretti et al, “A Mobile Delay-Tolerant Approach to Long-Term EnergyEfficient Underwater Sensor Networking,” In Proceedings of Wireless
Communications and Networking Conference (WCNC), Kowloon, pp. 2866 –
2871, 2007.
[33] K. D. Zoysa, C. Keppitiyagama, G.P. Seneviratne, and W. W. A. T. Shihan, “A
public transport system based sensor network for road surface condition
monitoring,” In Proceedings of the 2007 workshop on Networked systems for
developing regions (NSDR '07), ACM, New York, NY, USA, Article 9, 6 pages,
2007.
[34] D. Stavropoulos, G. Kazdaridis, T. Korakis, D. Katsaros, and L. Tassiulas,
“Demonstration of a vehicle-to-infrastructure (v2i) communication network
featuring heterogeneous sensors and delay tolerant network capabilities,” In
Proceedings 8th Intl ICST Conf on Testbeds and Research Infrastructure,
Development of Networks and Communities (TRIDENTCOM), 2012.
[35] T. Small and Z. J. Haas, “The Shared Wireless Infostation Model – A New Ad Hoc
Networking Paradigm (or Where there is a Whale, there is a Way),” In
Proceedings of the Ninth Annual International Conference on Mobile Computing
and Networking (MOBICOM), pp. 233–244, 2003.
[36] P. McDonald, D. Geraghty, I. Humphreys, S. Farrell, and V. Cahill, “Sensor
Network with Delay Tolerance (SeNDT),” In Proceedings of International
Conference on Computer Communications and Networks (ICCCN), pp. 13331338, August 2007.
[37] F. Gabrovšek, B. Grašič, M. Z. Božnar, P. Mlakar, M. Udén, and E. Davies, “Karst
show caves – how DTN technology as used in space assists automatic
environmental monitoring and tourist protection – experiment in Postojna
Cave”, Nat. Hazards Earth Syst. Sci., Vol 14, pp. 443-457, 2014.
[38] Internet Engineering Task Force’s (IETF) DTN workgroup, [Online]. Available:
152 References
http://datatracker.ietf.org/wg/dtn/charter/. (last accessed July 7th 2015)
[39] Internet Research Task Force’s (IRTF) DTN research group, [Online].
Available: https://irtf.org/dtnrg/. (last accessed July 7th 2015)
[40] The Consultative Committee for Space Data Systems (CCSDS), “CCSDS BUNDLE
PROTOCOL SPECIFICATION,” Draft Recommended Standard, CCSDS 734.2-R-2,
RED BOOK, Washington D.C., October 2013.
[41] V. Cerf et al., “Interplanetary Internet (IPN): Architectural Definition,” Internet
draft, draft-irtf-ipnrg-arch-00. txt, May 2001.
[42] S. Burleigh et al., “Delay-Tolerant Networking: an approach to Interplanetary
Internet,” IEEE Communications Magazine, Vol.41, Issue 6, pp. 128-136, July
2003.
[43] I. F. Akyildiz et al., “Interplanetary Internet: State-of-the-Art and Research
Challenges,” Computer Networks, Vol. 43, Issue 2, pp. 75-112, October 2003.
[44] G. Araniti, N. Bezirgiannidis, E. Birrane, I. Bisio, S. Burleigh, C. Caini, M.
Feldmann, M. Marchese, J. Segui, and K. Suzuki, "Contact graph routing in DTN
space networks: overview, enhancements and performance," Communications
Magazine, IEEE , vol.53, no.3, pp.38,46, March 2015.
[45] A. Lindgren, A. Doria, and O. Schelen, “Probabilistic routing in intermittently
connected networks,” SIGMOBILE Mobile Computing and Communications
Review, Volume 7, no3, pp.19–20, 2003.
[46] T. Berners-Lee, R. Fielding,and L. Masinter, “Uniform Resource Identifier (URI):
Generic Syntax,” RFC3986, January 2005.
[47] S. Symington, S. Farrell, H. Weiss, and P. Lovell, “Bundle Security Protocol
Specification,” RFC6257, May 2011.
[48] L. Wood, W. M. Eddy, and P. Holliday, ”A Bundle of Problems,” In Proceedings of
IEEE Aerospace Conference, pp. 1-17, Big Sky, Montana, March 2009.
[49] Y. Cao, Z. Sun, “Routing in Delay/Disruption Tolerant Networks: A Taxonomy,
Survey and Challenges,” Communications Surveys & Tutorials, IEEE, Vol 15,
Issue 2, pp. 654 – 677, 2013.
[50] Y. Zhu, B. Xu, X. Shi, and Y. Wang, “A survey of social-based routing in delay
tolerant networks: Positive and negative social effects,” IEEE Comm. Surveys
Tutorials, Vol. 15, Issue 1, pp. 387 - 401, 2012.
[51] J. Ott, “Towards more adaptive voice applications,” In Proceedings of the4th
International Symposium On Leveraging Applications of Formal Methods,
Verification and Validation (ISoLA 2010), Heraklion, Crete, Greece, October 1821, pp 355-366, 2010.
[52] S.-A. Lenas, S. C. Burleigh and V. Tsaoussidis, “Reliable Data Streaming over
Delay Tolerant Networks,” In Proceedings of the 10th international conference
on Wired/Wireless Internet Communication, Santorini, Greece, pp.358-365,
2012.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 153
[53] D. Henkel and T. X. Brown, “On Controlled Node Mobility in Delay-Tolerant
Networks of Unmanned Aerial Vehicles,” In International Symposium on
Advance Radio Technolgoies (ISART), March 7-9, 2006.
[54] W. Ivancic, “Delay/Disruption Tolerant Networking - Network Management
Requirements,” IETF Internet Draft, June 2009. [Online]. Available:
https://tools.ietf.org/html/draft-ivancic-dtnrg-network-management-reqs-00.
[55] F. Templin (Ed.) and S. Burleigh, “DTN Security Key Management Requirements and Design,” IETF Internet Draft, February 2015. [Online].
Available: https://tools.ietf.org/html/draft-templin-dtnskmreq-00.
[56] E. Birrane, S. Burleigh, N. Kaschc, “Analysis of the contact graph routing
algorithm: bounding interplanetary paths,” Acta Astronautica, Elsevier, Volume
75, pp. 108–119, June–July 2012.
[57] S. Burleigh, “Contact Graph Routing,” IETF Internet Draft, July 2010. [Online].
Available: http://tools.ietf.org/html/draft-burleigh-dtnrg-cgr-01S.
[58] A. Lindgren, A. Doria, E. Davies ,and S. Grasic, “Probabilistic Routing Protocol
for Intermittently Connected Networks,” IETF Internet Draft, May 2012.
[Online]. Available: https://tools.ietf.org/html/draft-irtf-dtnrg-prophet-10.
[59] N. Bezirgiannidis and V. Tsaoussidis, “Packet size and DTN transport service:
Evaluation on a DTN Testbed,” International Congress on Ultra Modern
Telecommunications and Control Systems 2010, Moscow, pp. 1198 – 1205,
October 2010.
[60] W.–B. Pottner, L. Wolf, F. Busching, and G. von Zengen, “Data elevators:
Applying the bundle protocol in Delay Tolerant Wireless Sensor Networks,” In
Proceedings of the 2012 IEEE 9th International Conference on Mobile Ad-Hoc
and Sensor Systems (MASS) (MASS '12), IEEE Computer Society, Washington,
DC, USA, pp. 218-226, 2012.
[61] N. Panchakarla, J. Ott, and G. Junior, “Delay-tolerant adaptive real-time
communication: a case study for voice,” In Proceedings of ExtremeCom’12,
2012.
[62] CISCO Corporation, “Cisco Visual
Methodology,” White paper, 2015.
Networking
Index:
Forecast
and
[63] M.Qin and R. Zimmermann, “Improving Mobile Ad-hoc Streaming Performance
through Adaptive Layer Selection with Scalable Video Coding,”In Proceedings
of 15th ACM Multimedia conference, pp. 717 – 726, 2007.
[64] C. Calafate, M. Malumbres and P. Manzoni, “Mitigating the impact of mobility
on H.264 real-time video streams using multiple paths,” Journal of.
Communication and Networks, Vol 6, Issue 4, pp. 387–396, 2004.
[65] V. Loscri, F. De Rango, and S. Marano, “Ad hoc on-demand multipath distance
vector routing (AOMDV) over a distributed TDMA MAC protocol for QoS
support in wireless ad hoc networks: Integration issues and performance
154 References
evaluation,” European Transactions on Telecommunications, Vol 18, Issue 2, pp.
141–156, 2006.
[66] K. Kambhatla et. al, “Wireless H.264 Video Quality Enhancement Through
Optimal Prioritized Packet Fragmentation,” IEEE Transactions on Multimedia,
14(5), pp.1480-1495, 2012.
[67] S. Mantzouratos, G. Gardikis, H. Koumaras and A. Kourtis, “Survey of Crosslayer Proposals for Video Streaming over Mobile Ad hoc Networks (MANETs),”
In Proceedings of International conference on Telecommunications and
Multimedia (TEMU ’12), pp.101-106, 2012.
[68] R. Patra and S. Nedevschi, “DTNLite: A Reliable Data Transfer Architecture for
Sensor Networks”, Tech. Rep. CS294–1, Berkeley 2003.
[69] C. Matarasso, D. Giggenbach, F. Hermanns and E. Lutz, “Multimedia Satellite
Communications Experiments to International Space Station,” International
Journal. of Satellite Communications, Vol. 20, pp. 333-345, 2002.
[70] S. Wang, D. Bian, X. Jie and Hong Liu, “The design and implementation of
channel simulator for broadband satellite communication,” International
Conference on Computer Science and Service System (CSSS), pp.4093-4096,
2011.
[71] A. Makovsky, P. Ilott and J. Taylor, “Mars Science Laboratory
Telecommunications System Design,” DESCANSO Design and Performance
Summary Series, Jet Propulsion Laboratory, California Institute of Technology,
November 2009.
[72] T. Liu and S. Nelakuditi, “Disruption-tolerant content-aware video streaming,”
In Proceedings of the 12th annual ACM international conference on Multimedia
2004, pp. 420–423, 2004.
[73] P. Tournoux, E. Lochin, J. Leguay and J. Lacan, “Robust streaming in delay
tolerant networks,” In Proceedings of. IEEE ICC, pp. 1–5, 2010.
[74] N. Kumar and J. Kim, “Probabilistic trust aware data replica placement strategy
for online video streaming applications in vehicular delay tolerant networks,”
Journal of Mathematical and Computer Modeling, Vol 58(1-2), pp 3–14, 2012.
[75] G. Karlsson, V. Lenders and M. May, “Delay-tolerant broadcasting,” In
Proceedings of. 2006 SIGCOMM workshop on Challenged networks, pp. 197–204,
2006.
[76] L. Keller, A. Le, B. Cici, H. Seferoglu, C. Fragouli and A. Markopoulou, “A.
Microcast: Cooperative Video Streaming on Smartphones,” In Proceedings of
the 10th international conference on Mobile systems, applications, and services
(ACM MobiSys ’12), pp. 57-70, 2012.
[77] Interplanetary Overlay Network (ION). Jet Propulsion Laboratory. California
Institute of Technology. CA. [Online]. Available: http://ion-dtn.sourceforge.net.
[78] DTN2,
DTN
Reference
implementation,
DTNRG,
[Online].
Available:
Towards a Future Internet Architecture: Protocols and Support Mechanisms 155
https://sites.google.com/site/dtnresgroup/home/code/dtn2documentation
(last visited: October 2016).
[79] M. Doering, S. Lahde, J. Morgenroth, and L. Wolf, “IBR-DTN: an efficient
implementation for embedded systems,” In Proceedings of ACM workshop on
Challenged networks (CHANTS), pp. 117–120, Sep. 2008.
[80] R. Sandau, H. Röser, A. Valenzuela, “Small Satellite Missions for Earth
Observation,” New Developments and Trends, Springer, 2010.
[81] C. Carlo, and R. Firrincieli, “DTN for LEO Satellite Communications,” Personal
Satellite Services, Springer, Berlin Heidelberg, pp. 186-198, 2011.
[82] E. J. Wyatt, S. C. Burleigh, R.M. Jones, J. L. Torgerson and S. S. Wissler,
“Disruption Tolerant Networking Flight Validation Experiment on NASA's
EPOXI Mission,” In Proceedings of The First International Conference on
Advances in Satellite and Space Communications (SPACOMM 2009), Colmar,
France, pp. 187-196, July 2009.
[83] W. Ivancic, W.M. Eddy, L. Wood, D. Stewart, C. Jackson, J. Northam and A. da
Silva Curiel, “Delay/Disruption-Tolerant Network Testing Using a LEO
Satellite,” In Proceedings of NASA Earth Science Technology Conference (ESTC
2008), University of Maryland, June 2008.
[84] G. Tyc, J. Tulip, D. Schulten, M. Krischke and M. Oxfort, “The rapideye mission
design,” Acta Astronautica, Vol. 56, no. 1-2, pp. 213–219, 2005.
[85] C Iacopino, P Palmer, N Policella, A Donati and A Brewer, “Self-Organizing MPS
for Dynamic EO Constellation Scenarios,” In Proceedings of: International
Workshop on Planning and Scheduling for Space (IWPSS 2013), At NASA-Ames,
USA.
[86] T. D. Rivers, “Small Satellites – Evolving Innovation for the Entire Market,” In
Proceedings of 31st Space Symposium, Technical Track, Colorado Springs,
Colorado, United States of America, April 13‐14, pp. 2015.
[87] G. L. Stephens et al., “The CloudSat Mission and the A Train,” Bull. Amer.
Meteorol. Soc., Vol. 83, pp. 1771–1790, 2002.
[88] M. Folkman, J. Pearlman, L. Liao, P. Jarecke, “EO1/Hyperion hyperspectral
imager design, development, characterization and prediction,” In Proceedings
of Hyperspectral Remote Sensing of the Land and Atmosphere (SPIE), W.L. Smith
and Y. Yasuoka (Eds.), Vol. 4151, pp. 40–51, 2001.
[89] R. Torres et al., “GMES Sentinel-1 Mission,” Remote Sensing of Environment, Vol
120, 9–24, 2012.
[90] “Spacecraft high rate data (hrd) radio frequency (rf) interface control
document (icd) to the direct-broadcast stations,” Goddard Space Flight Center
Greenbelt, Maryland, GSFC 429-03-02-24, May 10, 2004.
[91] Jorge Cantero Gómez, “Communication link design at 437.5 MHz for a
nanosatellite,” Diploma Thesis, Aalto University, June 2013.
156 References
[92] C.Wu, M. Donnelly, “Interface description document for eos aqua x-band direct
broadcast”, EOS AQUA PROJECT, NASA Goddard Space Flight Center Greenbelt,
Maryland, JUNE 2002.
[93] C. Brinton, E. Aryafar, S. Corda, S. Russo, R. Reinoso, and M. Chiang, “An
Intelligent Satellite Multicast and Caching Overlay for CDNs to Improve
Performance in Video Applications,” In Proceedings of 31st AIAA International
Communications Satellite Systems Conference, Florence, Italy, October 2013.
[94] E. Gill, P. Sundaramoorthy, J. Bouwmeester, B. Zand- Bergen and R. Reinhard,
“Formation flying within a constellation of nano-satellites: the QB50 mission,”
Acta Astronaut, 82(1), pp. 110–117, 2013.
[95] PACKET TELEMETRY,” CCSDS 102.0-B-5, Blue Book, November 2000.
[96] “TELECOMMAND”, CCSDS 201.0-B-3, Blue Book, June 2000.
[97] “AOS SPACE DATA LINK PROTOCOL”, CCSDS 732.0-B-2, Blue Book, July 2006.
[98] “CCSDS FILE DELIVERY PROTOCOL (CFDP),” CCSDS 727.0-B-2, Blue Book,
October 2002.
[99] B. Cohen, “The BitTorrent Protocol Specification,” January 2008. [Online].
Available: http://www.bittorrent.org/beps/bep_0003.html.
[100] S. Deering, “Host Extensions for IP Multicasting,” IETF RFC 1112, August
1989. [Online]. Available: https://tools.ietf.org/html/rfc1112.
[101] B. Adamson, C. Bormann, M. Handley, J. Macker, “NACK-Oriented Reliable
Multicast (NORM) Transport Protocol”, IETF RFC 5740, 2009. [Online].
Available: http://tools.ietf.org/rfc/rfc5740.txt.
[102] OFCOM, “Communications Market Report: UK – Smartphone Data Tables
(Adults),” [Online]. Available: http://stakeholders.ofcom.org.uk/marketdata-research/market-data/communications-market-reports/cmr11/, 2011.
[103] A. Sathiaseelan and J. Crowcroft, “LCD-Net: lowest cost denominator
networking,” ACM SIGCOMM CCR, Vol. 43, No. 2, pp. 52-57, April 2013.
[104] Open Wireless Movement, [Online]. Available: https://openwireless.org/
[105] A. Sathiaseelan, J. Crowcroft, M. Goulden, C. Greiffenhagen, R. Mortier, G.
Fairhurst and D. McAuley, “Public Access WiFi Service (PAWS),” Third Annual
Digital Economy All Hands Conference, Aberdeen, October 2012.
[106] FON Community, [Online]. Available: http://www.fon.com.
[107] Open Garden Community, [Online]. Available: https://opengarden.com.
[108] OpenSpark Community, [Online]. Available: https://open.sparknet.fi.
[109] Open-mesh, [Online]. Available: http://www.open-mesh.com.
[110] K. Carlberg, P. Gevros, and J. Crowcroft, “Lower than best effort: a design and
implementation,” In Workshop on Data communication in Latin America and
the Caribbean, San Diego, California, USA, pp.244-265, 2001.
Towards a Future Internet Architecture: Protocols and Support Mechanisms 157
[111] T. Chown, T. Ferrari, S. Leinen, R. Sabatino, N. Simar, and S. Venaas, “Less
than Best Effort: Application Scenarios and Experimental Results,” In
Proceedings of QoS-IP 2003, ser. Lecture Notes in Computer Science, no. 2601,
pp. 131–144, 2003.
[112] QBone Scavenger Service (QBSS), Internet2 QBone Initiative, [Online].
Available: http://www.internet2.edu/presentations/CampusWorkshopAtl
anta/Shalunov-QoS-Atlanta.pdf.
[113] M. Welzl and D. Ros, “A Survey of Lower-than-Best-Effort Transport
Protocols,”
IETF
RFC:
6297,
[Online].
Available:
https://tools.ietf.org/html/rfc6297/, 2011.
[114] S. Shalunov et al., “Low Extra Delay Background Transport (LEDBAT),” IETF
RFC: 6817, [Online]. Available: https://tools.ietf.org/html/rfc6817/, 2012.
[115] N. Kuhn et. al., “Less-than-Best-Effort capacity sharing over high BDP
networks with LEDBAT,” In Proceedings of IEEE 78th Vehicular Technology
Conference (VTC Fall), pp.1-5, September 2013.
[116] I. Psaras and L. Mamatas, “On Demand Connectivity Sharing: Queuing
management and load balancing for User-Provided Networks,” Computer
Networks, Vol. 55, No. 2 , pp.399-414, 2011.
[117] S.-A. Lenas and V. Tsaoussidis, “Traffic shaping for enabling less-than-best
effort services at the edges of broadband connections,” In Proceedings of the
2013 ACM MobiCom workshop on Lowest cost denominator networking for
universal access, Miami, Florida, USA, pp.41-46, 2013.
[118] D. Murray and T. Koziniec, “The state of enterprise network traffic in 2012,”
In the Proceedings of the 18th Asia-Pacific Conference on Communications
(APCC 2012), pp.179-184, 15-17 Oct. 2012.
[119] R. Sinha, C. Papadopoulos and J. Heidemann, “Internet packet size
distributions: some observations,” Technical report, October 2005.
[120] B. Dekeris, L. Narbutaite, and T. Adomkus, “Packet delay evaluation using
M/G/WFQ model,” In Proceedings of the 4th WSEAS International Conference
on Electronics, Hardware, Wireless and Optical Communications (EHAC'05),
February 13-15, 2005.
[121] A. Sathiaseelan et. Al, “A Feasibility Study of an In-the-Wild Experimental
Public Access WiFi Network,” In Proceedings of the Fifth ACM Symposium on
Computing for Development, ACM, New York, NY, USA, pp. 33-42, 2014.
[122] W. Lee and A. Fapojuwo, “Analysis and modeling of a campus wireless
network TCP/IP traffic,” Computer Networks, vol. 53, no. 15, pp. 2674–2687,
October 2009.
[123] M. Pustisek, I. Humar, and J. Bester, “Empirical Analysis and Modeling of
Peer-to-Peer Traffic Flows,” In Proceedings of IEEE MELECON 2008, pp. 169175, Sep. 2008.
158 References
[124] J. Rojas-Mora, T. Jim ́enez, and E. Altman, “Simulating flow level bandwidth
sharing with pareto distributed file sizes,” In Proceedings of VALUETOOLS ’11,
ICST, Brussels, Belgium, pp. 265–273, 2011.
[125] N. Basher, A. Mahanti, C. Williamson, and M. Arlitt, “A comparative analysis of
Web and Peer-to-Peer traffic,” In Proceedings of WWW, pp 287–296, 2008.
[126] J. K. Burbank, P. F. Chimento, B. K. Haberman and W. T. Kasch, “Key
challenges of military tactical networking and the elusive promise of MANET
technology,” IEEE Communications. Magazine, Vol 44, Issue 11, pp. 39–45,
2006.
[127] J. F. McGowan, “Video Technologies for Mars,” In Proceedings of the Second
International Convention of the Mars Society, 2000.
[128] C. Guo, “SRR: an O(1) time complexity packet scheduler for flows in multiservice packet networks,” In Proceedings of ACM SIGCOMM ’01, pp. 211-222,
San Diego, CA, Aug. 2001.
[129] J. Farber, “Network game traffic modeling,” In Proceedings of NETGAMES ’02:
1st workshop on Network and system support for games, New York, NY, USA,
pages 53–57, 2002.
[130] J.J. Lee and M. Gupta, “A New Traffic Model for Current User Web Browsing
Behavior,” technical report, Intel Corp., Santa Clara, Calif, 2007.
[131] I. Komnios, I. Alexiadis, N. Bezirgiannidis, S. Diamantopoulos, S.-A. Lenas, G.
Papastergiou and V. Tsaoussidis, “SPICE Testbed: A DTN Testbed for Satellite
and Space Communications”, In Proceedings of 9th International Conference
on Testbeds and Research Infrastructures for the Development of Networks and
Communities, TRIDENTCOM 2014, Guangzhou, China, May 5-7, 2014.
[132] Space Internetworking Center. http://spice-center.org.
[133] Lenas S.-A, “Worst-case analysis regarding the uniform application of packet
error rate across variable sized bundles,” [Online]. Available:
http://sotirislenas.info/documents/BSS-WCA.pdf.
[134] M. Ramadas, S. Burleigh, S. Farrel. “Licklider Transmission Protocol –
Specification,” Internet RFC 5326, September 2008.
[135] “LICKLIDER TRANSMISSION PROTOCOL (LTP) FOR CCSDS”, CCSDS 734.1-B1, Blue Book, May 2015.
[136] Cisco
Corporation,
“Fundamental
of
Digital
Video”,
[online]
http://www.cisco.com/en/US/docs/solutions/Enterprise/Video/pktvideoaa
g.pdf.
[137] VideoLan Media Player. http://www.videolan.org/vlc/index.html
[138] AGI,
Exton,
PA, USA.
Satellite
http://www.agi.com/products/stk/.
Toolkit.
[Online].
Available:
[139] S. McCanne, S. Floyd. Network Simulator. (1997) [Online]. Available:
Towards a Future Internet Architecture: Protocols and Support Mechanisms 159
http://www.isi.edu/nsnam/ns/.
[140] D. Vardalis, “DTN Agent for ns-2,” [Online]. Available: http://www.spicecenter.org/dtn-agent/.
[141] D. Vardalis, V. Tsaoussidis, “Exploiting the potential of DTN for energyefficient internetworking,” Journal of Systems and Software, Vol. 90, pp. 91–
103, 2014.
[142] “Netperf,” [Online]. Available: http://www.netperf.org/netperf/.
[143] “Project BISmark,” [Online]. Available: http://projectbismark.net.
[144] “Tcpdump,” [Online]. Available: http://www.tcpdump.org.
[145] “Tcptrace,” [Online]. Available: http://www.tcptrace.org.
[146] Cisco Corporation, “Cisco Visual Networking Index: Global Mobile Data
Traffic Forecast Update, 2012–2017,” White Paper, [Online], Available:
http://newsroom.cisco.com/documents/10157/1142732/Cisco_VNI_Mobile
_Data_Traffic_Forecast_2012_2017_white_paper.pdf.
[147] PAWS
Project
–
Phase
2,
[Online].
http://publicaccesswifi.org/2014/07/31/paws-2-launches/.
Available:
[148] S.A. Lenas, S. Dimitriou, F. Tsapeli, V. Tsaoussidis, “Queue- Management
Architecture for Delay Tolerant Networking,” In Proceedings of
Wired/Wireless Internet Communications Lecture Notes in Computer Science,
vol. 6649, pp. 470-482, 2011.