A BitTorrent Performance Evaluation Framework

In the last decade, file sharing systems have generally been dominated by P2P solutions. Whereas e-mail and HTTP have been the "killer apps" of the earlier Internet, a large percentage of the current Internet backbone traffic is BitTorrent traffic . BitTorrent has proven to be the perfect file sharing solution for a decentralized Internet, moving the burden from central servers to each individual station and maximizing network performance by enabling unused communication paths between clients. Although there have been extensive studies regarding the performance of the BitTorrent protocol and the impact of network and human factors on the overall transfer quality, there has been little interest in evaluating, comparing and analyzing current real world implementations. With hundreds of BitTorrent clients, each applying different algorithms and performance optimization techniques, we consider evaluating and comparing various implementations an important issue. In this paper, we present a BitTorrent performance evaluation framework that we are using to test and compare current real world BitTorrent implementations. The framework is fully automated and clients have been instrumented to output transfer status data and extensive logging information.