短相关特性对网络流量的测量与分析方法性能的影响与改进  

Impact of Short-range Dependence on the Performance of Network Analysis Methods

作　　者：林青家[1] 陈涤[1] 刘允才[1] 

机构地区：[1]山东大学信息科学与工程学院

出　　处：《电子测量与仪器学报》2006年第5期92-97,共6页Journal of Electronic Measurement and Instrumentation

基　　金：山东省自然科学基金(编号:Y2001G05);思科教育科研资助项目(编号:P0135006928)。

摘　　要：基于由FGN模型产生的具有LRD(long-rangedependence)特性的模拟流量和由FARIMA模型产生的同时具有LRD和SRD(short-rangedependence)特性的模拟流量,就SRD特性对网络流量特性分析方法性能的影响进行了研究,并提出了消除或减轻SRD特性影响的方法—聚类分析法。实验表明,SRD特性对网络流量特性分析方法的性能有很大的影响,上述方法可以有效地消除或减轻SRD特性影响,可以适用于实际网络流量的行为分析。LRD （ long-range dependence） and SRD （ short-range dependence） behaviors of traffic influence network modeling, service providing and traffic engineering. In order to understand this influence, several network analysis methods are applied to two kinds of simulated traces. One trace is generated by FGN model with LRD structure, another one is generated by FARIMA model with LRD and SRD simultaneously. Their performances are compared and they arrived at the conclusions： These methods exhibit good performance under the conditions with LRD structure, but they exhibit poor performance under that with LRD and SRD structure simultaneously. In order to reduce the impact of SRD, an aggregated method is presented. The results show that this method can eliminate or at least reduce the SRD structure. So it can be used as a tool for analyzing real traffic trace.

关 键 词：长相关 短相关 聚类分析 FARIMA模型 FGN模型 

分 类 号：TN915.01[电子电信—通信与信息系统]