基于端到端测量的链路状态概率快速推断方法  

An Approach to Fast Inferring Link State Probability Based on End-to-End Measurement

作　　者：赵佐[1] 蔡皖东[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710129

出　　处：《华南理工大学学报（自然科学版）》2010年第11期110-116,共7页Journal of South China University of Technology(Natural Science Edition)

基　　金：教育部博士点基金资助项目(200806990030);西北工业大学科技创新基金资助项目(2008KJ02028)

摘　　要：链路状态的概率分布作为先验知识对于推断链路性能状态的准确度起着重要作用.文中主要研究了在树形拓扑下基于端到端测量的内部链路状态概率推断问题,并将该问题定义为极大似然估计问题.采用乘积模型描述路径与链路的状态概率之间的关系,将链路状态概率的推断归结为路径状态概率的估计,提出了一种通过计算路径状态概率进而获得链路状态概率的方法,并将该方法用于仿真实验.结果表明,该方法具有较高的有效性和实用价值,能够准确有效地推断网络内部链路状态概率.As a kind of prior knowledge,link state probability distribution plays an important role in inferring the accuracy of link performance state.This paper deals with the inference of internal link state probability based on the end-to-end measurement in tree topology,and defines it as a maximum likelihood estimation problem.By using a product model to describe the relationship between the path and the link state probability and by estimating link state probability via the computation of path state probability,a new approach to the fast inference of link state probability is proposed.The approach is then applied to simulation experiments.The results indicate that the proposed approach is effective and practical in inferring the internal link state probability.

关 键 词：布尔网络断层扫描 链路状态概率推断 端到端测量 乘积模型 路径状态概率 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]