快速RMD法产生的自相似业务流的性质研究

Study of the Characters of Self-Similar Traffic Stream Generated by High-Speed RMD Method

传统的业务模型大多基于Poisson模型或其改进形式 ,假定业务突发长度显负指数分布 .近期真实网络流量分析表明很多信息源会产生在多时间尺度下具有自相似特性的信息流 .该性质显著地影响宽带网的流量控制及其排队分析 .合成可控制的自相似业务流是进行仿真的第一步 .实现了一个快速RMD算法的自相似流生成器 ;估计了产生的自相似流的Hurst参数 ;估计了两个或多个自相似流叠加后的流的Hurst参数 ;利用一种打乱算法成功地去掉了自相似流的长相关性 .

The conventional models are mostly based on Poisson models or their improved versions.It is assumed that the distribution of burst length is exponential. Recent traffic analyses of a lot of networks have shown that many traffic sources produce traffic streams that are self_similar over several time scales. This character has severe impact on flow control and queuing analysis in broadband networks. A self_similar traffic stream generator using a high_speed RMD algorthm has been implemented. The Hurst parameters are estimated for those self_similar traffic streams from the generator and several superposed self_similar traffic streams.The autocorrelation structure of the self_similar traffic is destroyed by shuffling algorithm.