IEEE 802.21 MEDIA INDEPENDENT HANDOVER
DCN: 21-10-0064-00-0000
Title: Providing Service Guarantees in Heterogeneous Wireless
Mesh Networks
Date Submitted: March, 17, 2010
Presented at IEEE 802.21 session #37 in Orlando, FL
Authors or Source(s):
Antonio de la Oliva and Albert Banchs (UC3M)
Abstract: This presentation shows how to abstract the capacity
concept of the specific wireless technologies
21-07-xxxx-00-0000
IEEE 802.21 presentation release statements
This document has been prepared to assist the IEEE 802.21 Working Group. It is
offered as a basis for discussion and is not binding on the contributing
individual(s) or organization(s). The material in this document is subject to
change in form and content after further study. The contributor(s) reserve(s)
the right to add, amend or withdraw material contained herein.
The contributor grants a free, irrevocable license to the IEEE to incorporate
material contained in this contribution, and any modifications thereof, in the
creation of an IEEE Standards publication; to copyright in the IEEE’s name
any IEEE Standards publication even though it may include portions of this
contribution; and at the IEEE’s sole discretion to permit others to reproduce in
whole or in part the resulting IEEE Standards publication. The contributor also
acknowledges and accepts that this contribution may be made public by IEEE
802.21.
The contributor is familiar with IEEE patent policy, as stated
outlined
in in
Section
Section
6 of
6.3the
of
the IEEE-SA
IEEE-SA
Standards
Standards
Board
Board
bylaws
Operations Manual
<http://standards.ieee.org/guides/opman/sect6.html#6.3> and
<http://standards.ieee.org/guides/bylaws/sect6-7.html#6>
and in
in
Understanding Patent Issues During IEEE Standards Development
http://standards.ieee.org/board/pat/guide.html>
http://standards.ieee.org/board/pat/faq.pdf>
21-07-xxxx-00-0000
Table of Contents
Objective
Carrier-grade services
Heterogeneous Wireless Mesh Networks
Resources in a wireless link
Capacity region
Linearization
Mapping to wireless technologies
Wireless LAN
WiMAX
Routing
Routing algorithm
Architecture view
Considerations and conclusions
3
Objective
Heterogeneous Wireless Mesh Networks (WLAN and WiMAX)
Set of requests with service requirements
Routing algorithm that satisfies all requests
CGW
CGW
CGW
CMN
CMN
CGW
CMN
WiMAX
CMN
CAP
R1
WLAN
CAP
CAP
R2
R3
4
Approach
Joint routing-MAC optimization
Routing algorithm
3rd part
1st part
2nd part
Resources
available
MAC algorithm
WLAN
Resources
configuration
MAC algorithm
WiMAX
Technologyindependent
Interface
(802.21 based)
Resources in wireless links
To meet the routing objectives, we need to know the resources
available in a wireless link
Resource availability in a wired link is straightforward
R  C
iL
i
Determining the resources available in a wireless link is much
more complex and depends on
Wireless technology (WiMAX, WLAN)
Number of stations
Sending rate of each station
Modulation scheme of the stations
Contention parameters
etc.
6
Capacity region of a wireless link
Capacity region
Wireless link with n stations
Throughput allocation {R1,…,RN}
Region of feasible allocations
Examples
R1
R1
R1
C
C1
C
WLAN with
homogeneous
rates
R2
C2
WLAN with
heterogeneous
rates
R2
R2
TDMA
MAC
protocol
7
Capacity region linearization
Objective
Definition of an interface to inform the upper layers
(routing) on resource availability
Simple interface based on a few parameters
Approach
Linearization of the capacity region
Link group model
R1
R1
c R
iL
R2
i
i
C
ci: cost of flow i
C: wireless capacity
R2
8
Capacity region parameters
Challenge
Computation of the parameters C, ci’s
Technology independent interface
Technology dependent algorithms
C, ci’s
Technologydependent
algorithm
Technology A
C, ci’s
Technologyindependent
Interface
Technologydependent
algorithm
Technology B
Mapping to wireless technologies
Wireless LAN
WiMAX
9
WLAN mapping
WLAN capacity region
Needed to compute the linearized capacity region
Set of throughput guarantees {R1,…,RN} that can be satisfied
The capacity region depends on the parameters setting
802.11e parameters: CWmin, CWmax, AIFS, TXOP
Goal: find the parameters setting that maximizes the
capacity region
Based on previous work we set
CWmin = CWmax = CWi
AIFS = DIFS
TXOP = constant
Goal
Find the optimal parameter configuration {CW1,…,CWN}
given the throughput requirements {R1,…,RN}
10
WLAN optimal configuration
Probability of transmission of a station:
2
i 
CWi  1
Relation between taus:
 i Ri

  i  wi 1
 j Rj
Rates can be expressed as an approximate function of 1 :
2
l (a 1  b 1 )
ri  
2
i c  d 1  e 1
To maximize the rates we take the derivative and isolate the
tau
ri
2
 1
 0  A 1  B 1  C  0
{CW1,…, CWN}
11
WLAN linearized capacity region
Capacity region
R1
set of allocations {R1, R2} feasible
with the optimal configuration
R2
Linearized capacity region
 Tangent to the capacity region
 Need to chose tanget point
R1
tanget point {T1, T2}
R2
12
WLAN tangent point
Prioritization
T1 > T2 Flow 1 is prioritized over flow 2 (and viceversa)
R1
R1
C1
C1
T1>T2
T2>T1
R2
R2
Compromise
R off flows (i.e. the ones with higher
Prioritizing better
modulation rate)
leads to a better overall performance
C
Worst off flow should not be overpenalized since they may
fair
correspond to critical routesproportionally
allocation
Well accepted tradeoff for this compromise: proportional
fairness
C2
C2
1
1
C2
R2
13
WLAN linearized capacity region
parameters
Computing the slope of the tangent hyperplane is difficult
Key Observation
We aim at
 ci ri  C
i
At the
tangent point
R1
C1
c r
i
From the above
C
i i
Ti
it follows
C2
R2
c r  c r
i i
i
i i
i
Ti
i.e. Σciri takes a minimum at Ti
From forcing the minimum, we obtain a linear system of
equations
  c j rj
0
j
Ri
ci
C   ci ri
i
Ti
Ti
14
WLAN homogeneous case
15
WLAN heterogeneous case
16
WLAN number of flows
Ri


Ri 1 1  
17
WLAN experimental results
18
WiMAX Mapping
WiMAX capacity region is already linear
Main challenge: compute wireless capacity C
2D to 1D Capacity Modeling (Downlink
Example)
R1
R1
R2
R2
C{MRBS,RS1 ,...RSN } (LG )  C{MRBS,RS1 ,... RSN }  COH {MRBS,RS1 ,... RSN }
 CU {MRBS,RS1 ,...RSN }  CR {MRBS,RS1 ,...RSN }
c { MRBS ,RS ,...RS } ( LG ) 
1
N
1
C{MRBS, RS1 ,... RSN } MCS { MRBS ,RS ,...RS
1
N}
19
Routing in heterogeneous mesh networks
Given
Network topology
Linearized capacity region for each wireless link
Source node of each flow
Througput request of each flow
Find
Admission control: which flow requests can be admitted and
which not
Routing: path of each admitted flow
Two routing options
MP: Multipath routing. A flow can be split between different
paths
SP: Single-path routing. A flow must be sent through a single
path
Note that multipath TCP extensions are currently being
standardized by the IETF
20
Multipath routing
Well known problem: multi-commodity flow problem
Linear programming formulation
minimize
r
ij
subject to
i, j

 c j rij  C
 j

 rij   rik
 j
k

 rij  Ri
 j
Capacity Constraint
Rate Conservation Constraint
Source Constraint
Can be solved using standard techniques

21
Single-path routing
Well known problem: unsplittable flow problem
Integer programming formulation
maximize
x
i
subject to
i
 cl  yi ,l Ri  C L
 lL il
 y 
yil

il

l
lN out
 N in
 xi   yi ,l

lsi

 xi , yil  {0,1}
Capacity Constraint
Rate Conservation Constraint
Source Constraint
Single-path Constraint
Can be solved using standard relaxation techniques
22
Routing experiments description
Scenario
Area of size 400m x 400m
Between 40 and 70 nodes
Multiple random topologies
35 topologies analyzed per experiment
Average performance and deviation
Topology generation
Hyacinth-Laca generator
Channels set up
Common Channel Set allocation
Rate between pairs of nodes
Throughput vs distance curves
23
Routing topology example
24
Routing Homogeneous mesh
25
Routing Heterogeneous mesh
26
Interference considerations
Mapping model assumption
Low radio interference between link groups
Provided by the self-configuration function
Experimental results
Floornet testbed
Interference with 802.11a very limited as long as channel
separation is enough
802.11a has enough channels for wide deployment
Extensions for interference
In some occasions we may suffer from interference
Model extensions to account for the probability that a noncolliding transmission fails due to interference
Ongoing work
27
Routing experimental results
Experimental results on routing
 Testbeds ~10 WLAN interfaces
 Routing without interaction with
MAC
Routing strategies
 Proposed approach
 Standard ETX
 Shortest path
(SP)
Preliminary results
 SP
 SP
outperforms ETX by x1,5 or x2
outperforms ShP by x3 or x4
Conclusions
 Experimental results
 More tests required
seem to confirm simulations
28
Experimental Results
Note: LG3 contains link at 6 Mbps
• We measure the losses
while adding flows of 1
Mbps from Server to
CAP
• We stop adding flows
when 5% losses
• Results
•
•
•
Dijkstra 4 Mbps
ETX 9 Mbps Mbps
Linear 16 Mbps
21-07-xxxx-00-0000
Summary and conclusions
Model definition
Definition of a model for available resources in a wireless
link
Technology independent interface
Linear region (allows solving optimality problems)
Mapping to wireless technologies
Optimal computation of the interface parameters
Results show the efficiency of the interface
Routing solution
Homogeneous and heterogeneous mesh
Multipath and single-path routing
It outperforms previous approaches by x2 or more
Ongoing work
DeIay, interference, DVB
30