自相似业务流复用特性分析

Statistical Multiplexing of Self-similar Traffic

现有研究将复用的自相似业务流Hurst参数值确定为各个业务流中最大的H参数值,与业务流的其他性质无关,这一结论用于网络设备的设计不利于网络资源的有效利用。本文采用简单近似估算,并用分形布朗运动模型生成自相似业务流,采用小波分析方法估计Hurst 参数值。实验结果表明,由于复用合成业务流的渐近自相似的本质,在可以观测的时间尺度范围内业务流的Hurst参数比这理论预测值小;在一定的序列长度下,复用流的Hurst参数的不仅和最大Hurst参数业务流有关,还受到其它业务流,特别是业务流的方差系数所表现出的短时突发性影响,因此对合成业务流的自相似参数具有重要的影响。

The network performance is influenced by the self-similarity of traffic, which is characterized by Hurst parameter. Self-similarity of multiplexing traffic as FBM (fractal brownian model) is investigated. Because the multiplexing self-similar traffic is asymptotic self-similar in nature, wavelet-based simulation demonstrates that the conventional multiplexing theory is conservative as not to utilize the network resources fully. Furthermore, within the FBM, most of self-similarity is contributed by the large variance traffic. This can be used to develop an effective method to diminish the self-similarity.