Student: Le Vu Hoang Minh Supervisor: Dr. Ben Leong UROP:U116150

1
Student Le Vu Hoang Minh
Supervisor Dr. Ben Leong UROPU116150
Implementation and Evaluation of Tit-for-Tat
Transfer Protocol
Introduction
Evaluation

• Peer-to-Peer
• improves data transfer by exploiting bandwidth
  between network participants.
• participants contribute computing power
  bandwidth, storage space, etc ? increase the
  capacity of the whole system.
• Examples of P2P systems Napster, Gnutella,
  Fasttrack, and the famous BitTorrent.
• Demand for file distribution
• Bandwidth expected to increase
• But file sizes are also increasing
• server becomes bottleneck
• P2P is a viable solution
• BitTorrents problems
• Throughput is not optimal
• Not Incentive-compatible
• Tit-For-Tat Transfer Protocol (TFTTP)
• Promises allow trading of blocks not already
  owned
• Correct incentives for peers to contribute their
  share

• Experiments on PlanetLab
• Choose 11 random nodes from PlanetLab one of
  which will act as the server. Run experiment on
  the set of nodes 3 times.
• File size is 100 MB
• Block size 256 KB. Each sector contains 5
  blocks.
• Compare promise vs. no promise
• Compare TFTTP vs. BitTorrent
• Results
• Promise vs. no Promise
• Download time 57 faster on average with
  promises
• Consistently higher throughput

Key Ideas

• Tit-For-Tat (TFT)
• File-Ring
• File divided into blocks
• Blocks organized as contiguous sectors

Conclusion

• Promises improve performance
• TFTTP has great potential to excel previous P2P
  architectures in file distribution.
• Future work
• Understand effect of Block size sector size
• Investigate algorithms for exchanging Peer-List,
  and deciding on which blocks to issue to clients
  and for the trades between clients
• Cluster of servers