基于网络延迟的检测点放置方法  

Network Monitor Placement Method Based on Network Delay

作　　者：石恒华[1] 何泾沙[2] 许鑫[1] 

机构地区：[1]北京工业大学计算机学院,北京100124 [2]北京工业大学软件学院,北京100124

出　　处：《北京工业大学学报》2009年第6期839-844,共6页Journal of Beijing University of Technology

基　　金：北京市自然科学基金资助项目(KZ200610005003)

摘　　要：为了解决网络延迟检测点的有效放置问题,根据网络节点之间的网络延迟,使用系统聚类法对网络节点进行聚类,并使用递归去除1度异常节点算法提高聚类效果.提出动态距离和最小算法,在聚类结果的类中选取对应的类中心,检测点放置在类中心上.仿真结果表明,系统聚类法中类平均法更适合对网络节点进行聚类,递归去除1度异常节点后能表现出更好的聚类效果,通过动态距离和最小算法能找到放置检测点的最佳位置.使用系统聚类法中的类平均法、递归去除1度异常节点算法、动态距离和最小算法,能有效地解决网络检测点的放置问题.For network measurement, the efficient placement of measurement nodes is the key of the network measurement. Basing on the value of the network distance between the nodes, the hierarchical clustering method set are selected to cluster the network nodes. The algorithm of recursively eliminating one out-degree singular nodes improves the clustering effect. A dynamical least distance summation algorithm is put forward to select cluster centering on the cluster result, and the cluster centers are chosen as the monitors. The simulation results show that average linkage algorithm of the hierarchical clustering method set can better cluster the network nodes, and the two proposed algorithms can improve the clustering effect better and select the right cluster centers. The problem of the network monitor placement can be effectively solved with the average linkage algorithm of the hierarchical clustering method set, the algorithm of recursively eliminating one out-degree singular nodes and the dynamical least linkage summation algorithm.

关 键 词：网络拓扑 聚类 网络延迟 异常节点 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]