大规模网络中BitTorrent流行为分析

Analysis of BitTorrent flow behavior on large-scale networks

在使用基于特征串方法准确采集国内最流行的P2P应用——BitTorrent应用流量的基础上,研究了BitTorrent应用的流长、流持续时间、流速以及端结点传输的流量、连接数等测度的分布,并分析了各分布中的异常情况.分析结果表明:BitTorrent应用流量已占网络总流量的60%;BitTorrent平均流长超过总体TCP流长的20倍;流长、流持续时间均服从Weibull分布,且二者行为相似;BitTorrent平均流速低于总体TCP流速,但随着流长的增长流速增快;BitTorrent网络有很强的不平衡性,约1%的结点贡献了80%的流量和连接数.

A trace of BitTorrent application traffic, which is the most popular P2P（peer-to-peer） application in China, is accurately captured using payload-based method. Based on this trace, the distributions of BitTorrent flow length, flow duration, flow rate, the nodes＇ traffic volume and number of connections are studied, and the unusual circumstances of these distributions are analyzed. The re- suits show that BitTorrent application traffic account for 60% of the total traffic volume. The average flow length of BitTorrent application is 20 times longer than that of overall TCP flow. The distributions of the flow length and flow duration approximately accord with the Weibull distributions, and their behaviors are similar. The average flow rate is less than overall TCP flow; however with the flow length increases, the flow rate becomes faster. The BitTorrent network presents a strong imbalance, about 1% of the nodes contribute 80% of the total volume and connections.