Kai RenCarnegie Mellon UniversityBitTyrant: Exploit AltruismBitTorrent Maintain an active set based on rate Tit-for-tat strategy Favors peers who sends back data rapidly Optimistic unchoking Randomly choose a small number of peers Help bootstrap of new peersObservation Reciprocation from Q to P The rate of data sent from P to Q > The rate of data sent from others to QBitTyrant: exploit altruism Strategy Maximize reciprocation bandwidth Maximize number of reciprocating peers Deviate from equal split Algorithm For p, dp–download rate, up– upload rate Rank peers by dp/ up Estimate dpby observed rate Dynamically adjust upby multiplicative factorsQuestions and Discussions Evaluation: Why ignore files larger than 1GB? Why only used 5MB files for experiment? Any metrics about fairness? Long-term performance Do nowadays popular P2P clients adopt the strategy from BitTyrant?Other Discussions Membership and Authentication in P2P Proof –carrying authorization With centralized server? P2P and ISPs Cooperation between P2P system and ISPs?EndObservation High upload capacity peers can enhance performance by using strategyBad Cases Low benefit for low capacity peers in multiple swarms cases Lengthy bootstrapping period Will it cause new users to contribute more? Peering relationship not stable Churn is
View Full Document