BitTorrent is an Auction
Incentives play a crucial role in BitTorrent. By motivating users to upload to other peers, all members of the swarm achieve faster downloads. Though it was once believed that BitTorrent was impervious to cheating, a recent empirical study showed that it is possible to strategically manipulate BitTorrent's unchoking algorithm. This strategy allows peers to upload less without increasing their download time.
Because the standard tit-for-tat model does not capture this behavior, we propose an auction-based model to better understand and improve upon BitTorrent's incentives. This model captures known performance-improving strategies, and also reveals new strategies. For example, we demonstrate that BitTorrent peers have incentive to intelligently under-report what pieces of the file they have.
We have implemented and evaluated a strategy-proof BitTorrent client in which peers reward one another with proportional shares of bandwidth. In experimental evaluation (on PlanetLab, on a local cluster, and in live swarms), our PropShare client achieves faster downloads than BitTorrent and BitTyrant clients, and our under-reporting strategy allows peers to maintain neighbor interest longer.
Papers
BitTorrent is an Auction: Analyzing and Improving BitTorrent's Incentives (pdf)
Dave Levin, Katrina LaCurts, Neil Spring, Bobby Bhattacharjee
SIGCOMM 2008
Downloads
Source code for PropShare BitTorrent client
People
Dave Levin
Katrina LaCurts
Neil Spring
Bobby Bhattacharjee