自相似业务网络性能分析

On performance analysis of self-similar traffic

最近10多年的大量研究结果表明,网络业务具有普遍的自相似特性。自相似特性对网络的性能分析和设计具有广泛的影响。自相似业务的性能分析问题是一个没有很好解决的问题。提出了自相似业务环境下的分形界到达过程,研究了基于分形界到达过程的自相似业务漏桶参数的优化方法,给出了时延约束条件下漏桶速率和桶深的计算公式。最后研究了网络演算在确保服务性能分析中的建模方法,推导出了基于网络演算的端到端时延界、队列长度界和有效带宽的计算公式。通过具体的应用实例,验证了分析结果的正确性和有效性。

The results of extensive studies in the last decade indicate that network traffic has the ubiquitous characteristic of self-similarity. Self-similarity has great impact on the desigu and performance analysis of networks. The problem of performance analysis of self-similar traffic is an unsolved problem. An envelope process called fractional bounded arrival process （FBAP） is proposed for self-similar traffic representation. The optimal method of parameters of token bucket is studied based on FBAP traffic. The formal of the rate and depth of the token bucket based on delay constrained are derived, A model for the performance analysis of guarantee service based on network calculus is proposed. The maximum end-to-end delay bound, the maximum backlog bound and the effective bandwidth of guaranteed service are derived. The validity and effectiveness of the analytic results are verified with an application example.