On the Impact of Practical P2P Incentive Mechanisms on User Behavior

Abstract

In this paper we report on the results of a large-scale measurement study of two popular peer-topeer systems, namely BitTorrent and eMule, that use practical and lightweight incentive mechanisms to encourage cooperation between users. We focus on identifying the strategic behavior of users in response to those incentive mechanisms. Our results illustrate a gap between what system designers and researchers expect from users in reaction to an incentive mechanism, and how users react to those incentives. In particular, we observe that the majority of BitTorrent users appear to cooperate well, despite the existence of known ways to tamper with the incentive mechanism, users engaging in behavior that could be regarded as cheating comprised only around 10% of BitTorrent’s population. That is, although we know that users can easily cheat, they actually do not currently appear to cheat at a large enough scale. In the eMule system, we identify several distinct classes of users based on their behavior. A large fraction of users appears to perceive cooperation as a good strategy, and openly share all the files they obtained. Other users engage in more subtle strategic choices, by actively optimizing the number and types of files they share in order to improve their standing in eMule’s waiting queues; they tend to remove files for which downloading is complete and keep a limited total volume of files shared.