基于层次分解的网络链路时延分布快速推测算法.pdfVIP

第35 卷第8 期 电 子 与 信 息 学 报 Vol.35No.8 2013 年8 月 Journal of Electronics Information Technology Aug. 2013 基于层次分解的网络链路时延分布快速推测算法 杨京礼* 孙 超 姜守达 魏长安 (哈尔滨工业大学自动化测试与控制研究所 哈尔滨 150080) 摘 要：为提高离散时延模型下的网络链路时延分布的测量速度，该文提出一种基于层次分解的网络链路时延分布 快速推测算法。按照树型网络拓扑的层次对端到端路径时延进行子树分解，以子树作为链路时延分布计算的基本单 元，减少端到端路径时延分解到链路时延后造成的冗余计算，缩短链路时延分布的计算时间。仿真实验结果表明， 该文提出的算法能够在不损失测量精度的条件下，有效缩短测量总时间，提高链路时延分布的测量速度。 关键词：网络层析成像；链路时延分布；层次分解；期望最大化算法 中图分类号：TP393 文献标识码： A 文章编号：1009-5896(2013)08-2005-08 DOI: 10.3724/SP.J.1146.2012.01455 A Fast Link Delay Distribution Inference Algorithm Based on Hierarchy Decomposition Yang Jing-li Sun Chao Jiang Shou-da Wei Chang-an (Department of Automatic Test and Control, Harbin Institute of Technology, Harbin 150080, China) Abstract: In order to speed up the inference algorithm of the link delay distribution based on discrete delay model, a fast inference algorithm based on hierarchy decomposition is proposed. This algorithm decomposes the end-to-end path delay into subtree units by the levels of the topology, and calculates the link delay distribution based on those subtree units. Through the reducing of the redundancy decomposition process of the end-to-end path delay, it can speed up the process of the inference of the link delay distribution. Simulation results show that this algorithm can improve the speed of the link delay distribution without loss of the accuracy. Key words: Network tomography; Link delay distribution; Hierarchy decomposition; Expectation Maximization (EM) algorithm 1 引言 (Computerized Tomography, CT)思想引入到网络 近年来，随着计算机网