*** 1 - Networking Research Lab @ SYSU

Presented by: Su Yingbin
Outline
 Introduction
 SocialSwam Design
 Notations
 Algorithms
 Evaluation
 Conclusion
Tit-for-tat as incentive to upload
 Want to encourage all peers to contribute
 Peer A said to choke peer B if it (A) decides not to
upload to B
 Each peer (say A) unchokes at most 4 interested
peers at any time
 The three with the largest upload rates to A
 Where the tit-for-tat comes in
 Another randomly chosen (Optimistic Unchoke)
 To periodically look for better choices
Typical BitTorrent incentives create inefficiencies
 Clients typically avoid increasing the number of
unchoke slots
 Bandwidth reserved to peers won’t actually be used
totally.
 Social hubs can’t receive the highest priority in
receiving file
Karame et al. show that combining locally optimal
solutions of the smaller social teams would give a
globally optimal solution for the entire social network.
Just work as a team!
SocialSwam Design Goal
 Maximize collaboration between social peers
 Maintain game-based techniques to encourage the
cooperation of non-social peers
SocialSwarm
Interaction Overview
1.
Retrieve social peers and
non-social peers from
tracker
2.
Identifies Bob’s social
peers
3.
Coordinates chunk
collection with them
4. Altruistically shares
bandwidth with them
5.
Interact with each other as
well as standard BitTorrent
clients
How ?
 How to identify social peers and non-social peers ?
 Social Distance
 How to collaborate with each other among a social
group as well as non-social peers ?
 Adaptive Bandwidth Allocation
 Chunk Prioritization
 Optimistic Unchoke Candidate Selection
Notations
Altruism Between Direct Social Peers
•I(a, b) is the number of reciprocal interactions a has had within a given time
window with b
•I(a, all) is the number of reciprocal interactions a has had with all of its peers
during
the same window of time.
•A(a, b) represents the proportional willingness that a peer a has to share resources
with each of its direct peers
Approximating SocialDistance Between Indirect Peers
-------- direct peers
Peers beyond this value
are considered as nonsocial
Notations
Overall Rarity for Each Given Chunk
Social Rarity for Each Given Chunk
Non-social Rarity for Each Given Chunk
The “gather-and-share” Technique
 From the social group perspective
 When the average social rarity for all chunks is high,
allocate more bandwidth for non-social peers.
 As the average social rarity for all chunks decreasing,
allocate more bandwidth for social peers.
 Average social rarity for all chunks:
 Maximum percentage of bandwidth allocated to social
peers:
The “gather-and-share” Technique
 From the social individual perspective
combines the social, non-social, and
overall rarities to form a combined
 Chunk prioritization
weighted rarity for each given chunk
target a peer with the largest
group of rare chunks at each
time interval ti
 Optimistic Unchoke Candidate Selection
SocialSwarm in a Nutshell
Social Network Data Set
 500 nodes with their interactions – Wall Postings –
extracted from Facebook
 Each pair of reciprocal postings is considered a single
interaction.
 Interactions are used to determine the direct level of
altruism between Facebook users.
 Beyond MaxSocialDistance are considered as non-
social peers
Baseline Test Parameters
Comparison of Basic Download Time
Client Download Rate Comparison
Chunk Rarity Reduction Comparison
Effect of File Size on Peer Throughput
Effect of Maximum SocialDistance on Peer Throughput
Effect of Additional Seed Capacity
Bandwidth Contribution and Unchoke Slot Allocation
Conclusion
 Typical incentives create inefficiencies
 SocialSwarm exploits SocialDistance to reduce this
inefficiencies
 The “gather-and-share” technique achieve better
performance