Internet中Peer-to-Peer应用流量测量与分析

Peer to Peer网络(简称p2p)作为一种新型的覆盖网络引起了越来越多研究者的兴趣。本文介绍了在我国进行的骨干互联网上p2p网络流量测量。与现有国外研究不同,本文的数据来源于核心路由器,因此克服了它们的缺陷。其研究集中在汇聚流中的3个周期性尖峰群、不同主机发送或接收流量的重尾分布、p2p流量的长相关特性以及提出了ADTE的估计方法来区分信令和数据流量。本文的研究也显示出Napster在p2p流中占大部分,这暗示着超级节点和阶层式拓扑较纯p2p结构潜在的优势。同时,观察到在我国p2p的流量仅占Internet总流量的1%弱,这个值跟国外的数据有很大区别。我们分析了其中的原因并希望该结论为我国p2p软件的发展提供参考。

Characterizing Peer-to-Peer Traffic Across Internet

As a new-type popular overlay network, the peer-to-peer network ( p2p for short) has drawn more and more attentions in research community. This paper, to the best of our knowledge, for the first time, launches a peer-to-peer network traffic measurement across the Internet backbone in China. Different from the existing studies, the data are derived from core routers on the Internet and overcome the limitations of those studies. This study focuses on the three periodic peak value groups traffic characteristics on flow statistics in the aggregation traffic, the heavy-tailed property in the distribution of traffic transmitted or received by different hosts and the long range dependence (LRD) property in p2p traffic. Moreover the paper proposes a novel data mining approach, ADTE to differentiate signaling and data traffic. It also shows that Napster, a hybrid p2p system, dominates p2p traffic which indicates some potential advantages of super-nodes and hierachical topology over pure p2p structure. In addition, it suggests that the aggregation of p2p traffic is less than 1% of the total Internet traffic, much different from the observations in other studies. The paper analyzes the possible reason. We wish that the conclusion could be a good guideline for the software developers.