网络流量的小尺度特性分析  

Small-time Scaling Behaviors of Network Traffic

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,高级工程师山东济南250100

出　　处：《电子测量与仪器学报》2007年第3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);思科教育科研资助计划(编号:PO135006928)

摘　　要：网络流量的尺度特性对于网络模型的建立、网络特性的提高以及网络的设计均有重大的影响,已成为目前网络行为研究的一个焦点。本文基于小波的多尺度分析方法,对网络流量的小尺度(1——100ms)特性进行了分析,结果表明,网络流量在小尺度上表现出非平稳、非LRD和多分形的特征。将流量按流大小和流密度进行分类,分类后的研究结果表明,流的密度对流量的小尺度特性影响最大。Scaling behaviors of network traffic influence network modeling, service providing and traffic engineering. So it has been the focus of many researches. In order to understand the causes of small-time scaling phenomena, an extensive wavelet analysis was performed to TCP flows. The results show that for a majority of the traces, the second-order scaling indices at small-time scales （1 -100ms） are fairly close to 0. 5. It indicates that traffic fluctuations at these time scales are nearly uncorrelated. In addition, the traces manifest non-stationary, non- LRD and muhi-fractal behaviors at small-time scales. The flow composition of the traffic is analyzed along with two dimension-flow byte contribution and flow density. And the results show that the flow density is the correlationcausing factor in small-time scales.

关 键 词：小尺度 多分形 密集流 小波 

分 类 号：TN915.01[电子电信—通信与信息系统]