Achieving Fairness in BitTorrent

By harnessing the collective storage capacity and bandwidth of Internet users, peer-to-peer le sharing systems have proved invaluable for storage and distribution of les. However, such systems have a clear and fundamental problem: users have an obvious incentive to download, but no inherent incentive to upload. In naive le sharing protocols, this results in a prisoner's dilemma: the pareto-optimal outcome is for everyone to upload as much as they can, and download as much as they want, resulting in high download rates for all users and the costs involved in uploading evenly distributed across peers; however, any single agent stands to bene t from defection. By reducing uploads, an agent can both improve his download rate and reduce his total upload. This being the case, the Nash equilibrium occurs where users share little, or nothing at all. BitTorrent, a popular and e ective peer-to-peer le sharing system, attempts to increase incentives to share in order to achieve, or at least approach, the pareto-optimal outcome. However, analysis and practice have revealed aws with the existing implementation of BitTorrent, and have suggested methods and strategies which a peer can use to minimize his own uploads without signi cant reprecussions. We introduce and motivate the BitTorrent protocol, and introduce some of the models that have been used to analyse it. We then examine some of the aws that have been discovered and exploited. Finally, we introduce some of the strategies and methods that have been proposed to combat the free rider problem, leading users to a more appealing and socially equitable equilibrium.