2005 · 164 citations · 11 references
ComputingComputer ScienceComputer SystemsComputer NetworksDigital CommunicationDistributed SystemsCommunication NetworksNetwork Performance AnalysisPeer-to-peer NetworkingPeer-to-peer NetworksPeer-to-peer SystemsResource ProvisioningBandwidthBandwidth UtilizationContent DistributionLink UtilizationLink Utilization AnalysisPeer-to-peer Protocols
BitTorrent is a popular, scalable peer‑to‑peer file distribution system that efficiently delivers large files without overloading the origin server, yet its upload utilization, fairness, and mechanism choices in varied settings remain uncertain. The study aims to deconstruct BitTorrent and evaluate how its core mechanisms affect overall performance across diverse workloads. The authors use a simulation framework to assess peer link utilization, download time, and fairness metrics under various workloads. BitTorrent achieves near‑optimal uplink utilization and download times except in extreme cases, but low‑bandwidth peers tend to download more than they upload when high‑bandwidth peers are present.
ABSTRACT In recent years, BitTorrent has emerged as a very popularand scalable peer-to-peer file distribution mechanism. It hasbeen successful at distributing large files quickly and ef-ficiently without overwhelming the capacity of the originserver.EarlymeasurementstudiesverifiedthatBitTorrentachievesexcellentuploadutilization,butraisedseveralquestionscon-cerningutilizationinsettingsotherthanthosemeasured,fair-ness, and the choice of BitTorrent’s mechanisms. In this pa-per, we present a simulation-based study of BitTorrent. Ourgoal is to deconstruct the system and evaluate the impact ofits core mechanisms, both individually and in combination,onoverallsystemperformanceunderavarietyofworkloads.Our evaluation focuses on several important metrics, includ-ing peer link utilization, file download time, and fairnessamongst peers in terms of volume of content served.OurresultsconfirmthatBitTorrentperformsnear-optimallyin terms of uplink bandwidth utilization, and download timeexcept under certain extreme conditions. On fairness, how-ever, our work shows that low bandwidth peers systemati-cally download more than they upload to the network whenhigh bandwidth peers are present. We find that the
11
Rudolf Ahlswede, Ning Cai, Shuo Li et al. · IEEE Transactions on Information Theory · 2000 · 7.9K citations
Incentives Build Robustness in Bit-Torrent
Bram Cohen · 2003 · 2.7K citations
Eytan Adar, Bernardo A. Huberman · First Monday · 2000 · 1.5K citations · Full text