基于小波变换的实际网络流量刻画

Depict of the Real Network Traffic Base on Wavelet Transform

大量研究结果表明实际网络流量具有明显的尺度特性,在大尺度上表现出自相似,在小尺度上表现出多重分形。多重分形为刻画流量在小尺度上的奇异性提供了良好的数学框架,而小波变换对具有长程依赖性的流量起到了去相关的作用,因此有必要利用小波技术来研究多重分形。同时网络流量的多尺度特性也为研究人员提供了新的方法来探讨流量本质特征。本文基于小波技术研究实际网络流量,首先从全局尺度和局部尺度上分析流量特征,确定产生分形的时间。然后比较多媒体流量和数据型流量在不同尺度下所表现出的性能,并且给出了产生这种现象的原因。

Many research studies have proposed scaling of real network traffic has a notable effect on the network performance： the self-similar phenomena over large scale, the multi-fractal phenomena over small scale. As multi-fractal providing the upstanding frame of mathematics to depict the singularity of traffic over small scale, and the wavelet transform taking effect on un-correlation to the traffic which is the long range dependent, it is necessary to research the multi-fractal model base on wavelet technique. The multi-scale provides the new method to study the characteristic of traffic for researchers. The paper studied the real network traffic based on the wavelet technique, and analyzed the characteristic from global scale and local scale at first, then it could fix the time which generated fractal. At last, it found the difference of performance between multimedia traffic and data traffic in the different scale, and the reason of the phenomenon was given.