A Case Study in Modeling Large-scale Peer-to-peer File-sharing Networks Using Discrete-event Simulation

Peer-to-Peer file-sharing networks are garnering a significant share of home broadband networks. Since 2001, BitTorrent has been one of the most popular protocols for file-sharing. What makes BitTorrent so unique compared to past efforts is that it provides a builtin mechanism to ensure the fair distribution of content and prevents selfishness on the part of peers using game theoretical “tit-for-tat” piece distribution algorithms. While this is good for home consumers, Internet service providers are not pleased with how BitTorrent has overloaded their networks. To better understand the BitTorrent protocol, we have created a detailed simulation model that is able to scale to 10’s of thousands of peers using commodity hardware. In this paper, we describe our model design and implementation approach as well as present a validation and performance study of the model. A key result from this study is that our model’s memory footprint ranges between 67 KB to 2.3 MB per client peer depending on the simulation configuration parameters. This is 30 to 1000 times less memory than required by the operational BitTorrent software.