基于小波分析的WiMax网络业务流长相关性  

Wavelet analysis-based long range dependence in WiMax traffic

作　　者：胡永东[1,2,3] 吴国新[1,2] 徐逸卿[1,2,3] 

机构地区：[1]东南大学计算机科学与工程学院,南京211189 [2]东南大学计算机网络和信息集成教育部重点实验室,南京211189 [3]南京林业大学信息科学技术学院,南京210037

出　　处：《东南大学学报（自然科学版）》2013年第1期1-5,共5页Journal of Southeast University：Natural Science Edition

基　　金：江苏省自然科学基金资助项目(BK2011335)

摘　　要：为了精确有效地检测WiMax网络业务流的长相关性,利用自相似随机过程和小波变换都具有尺度不变性的特性,建立了表征长相关性的Hurst参数和小波系数的统计量.研究了业务流长相关性的小波估计法,反复选择小波消失矩和优化尺度参数的范围,减小这些因素对检测结果的影响.利用ON/OFF模型仿真WiMax网络中5种不同类型业务流,然后使用小波估计法对trace文件进行分析.检测分析结果表明:实时业务流的长相关性比较小,非实时业务流的长相关性相对较大;随着网络负载的增大,业务流的长相关性会略有变化,聚合业务流的长相关性的检测值与理论值基本吻合.In order to detect long-range dependence （LRD） in the WiMax network traffic accurately and efficiently, using the scale invariance property of self-similar stochastic process and the wavelet transform, the statistics of LRD Hurst parameters and wavelet coefficients are established. Then, the wavelet analysis algorithm of LRD in network traffic is studied. Repeated selection of wavelet vanis- hing moments and the optimization of the range of the scale parameter can reduce the impact on the test results. By the ON/OFF model, five different types of traffic in WiMax networks are simulated. And then the wavelet analysis method is used to analyze the trace file. The test results show that LRD of the real-time traffic is small, but LRD of non-real-time traffic is relatively large. As the network load increases, LRD of network traffic may change slightly. The detection value of LRD of aggregate flow is basically consistent with the theoretical value.

关 键 词：WiMax业务流 长相关性 HURST参数 小波分析 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]