| BitTorrent systems: Content distribution |
![]() "PlanetLab is a convenient platform for testing peer-to-peer protocols and services." Arnaud Legout, Researcher, INIRA Participating organizations: INRIA (France), Poznan University of Technology (Poland), UCLA (USA) Peer-to-peer content distribution systems enjoy great popularity and are now gaining momentum as a means of disseminating large amounts of data over the Internet. Researchers from this multinational team have been studying the performance and robustness of BitTorrent, one of the most popular peer-to-peer protocols. They first conducted a first set of experiments in late 2006 to understand the properties of BitTorrent's peer selection algorithm. These experiments were performed on 40 PlanetLab nodes using an instrumented BitTorrent client. The researchers demonstrated the importance of good provisioning for the initial seed, the first peer that makes content available, on the performance of BitTorrent. In particular, they showed that the effectiveness of BitTorrent's sharing incentives depends on the upload capacity of the initial seed. The results of this study have been published at ACM SIGMETRICS 2007. In a second set of experiments, conducted in late 2007, the researchers evaluated the optimal piece size in BitTorrent, a critical parameter that had been largely overlooked by previous research efforts. They conducted experiments with varying piece sizes on a controlled BitTorrent testbed built on top of 40 PlanetLab nodes. The results demonstrated that this parameter is indeed critical, as it determines the degree of parallelism in the system, and discussed the optimal piece sizes for distributing small and large content. The researchers also pinpointed a related design tradeoff, and explained how BitTorrent's choice of dividing pieces into subpieces attempts to address it. This study has been published at IPTPS 2008. For further details, as well as the instrumented BitTorrent client used in the experiments, please visit the project webpage. Related papers: N. Liogkas, A. Legout, E. Kohler, L. Zhang. Clustering and Sharing Incentives in BitTorrent Systems. In Proceedings of ACM SIGMETRICS 2007. San Diego (USA): June 2007. P. Marciniak, N. Liogkas, A. Legout, E. Kohler. Small Is Not Always Beautiful. In Proceedings of IPTPS'2008. Tampa Bay (USA): February 2008.
|