对等网络环境下扩展树的统计模型及其性能分析

Statistical analysis of spanning trees in peer-to-peer networks

作为传统组播技术的取代方案,应用层组播获得了广泛研究。该文定义了稳定性因子和标准化网络延迟来量化和评估对等网络环境中拥塞对扩展树性能的影响。提出一种随机链路模型来描述扩展树中的拥塞,并且引入相关度因子来表达相邻链路之间的相关性。数字仿真结果表明:即使在高链路拥塞概率的情况下,增大链路之间的相关性,扩展树也会变得更稳定。该统计模型也适用于指导设计应用层组播的拥塞控制协议。

Application-layer multicast (ALM) has gained popularity as an alternative of traditional multicasting. This article defines a stability factor and a standard network delay to quantify and evaluate the impact of congestion on spanning trees in Peer-to-Peer (P2P) networks. The article first develops a simple statistical link model which describes congestion occurring in spanning trees and then introduces a dependency-degree factor which expresses the dependence between adjacent links. Simulation results show that even in case of high congestion probability, the stability of spanning trees can be significantly enhanced by improving the dependence of adjacent links. The statistical model not only provides a framework to compare existing ALM algorithms, but can also be applied to ALM network performance analysis.