自相似业务量下网络的排队性能分析

在研究通信网的网络性能、流量控制和资源配置时，业务源模型是其中一个十分重要的组成部分。近年来的研究发现数据网络中的业务量呈自相似特征，传统的能够精确描述电话网络的马尔可夫模型不再适用。业务量的自相似特征显著影响网络的流量控制与排队分析，已经引起人们的极大重视。采用G/M/1排队模型对分组长度可变的网络的排队性能进行了分析和仿真，其中自相似业务量是通过叠加大量独立的到达间隔为Pareto重尾分布的ON/OFF源来生成的。仿真结果表明，自相似业务量导致网络的排队性能劣化，这与有关文献对ATM交换的分析结果一致。

Queuing Performance Analysis with Self-similar Network Traffic

When studying the network performance,flow control and resource provisioning of communication networks,traffic model plays a very important role.The recent studies show that the date network traffic is self-similar,so the Markovian model,which describes telephone networks accurately,is not suitable for date networks.The self-similarity of the network traffic has severe impact on flow control and queuing analysis in date networks,therefore it has received significant attention.In this paper,G/M/1 queuing model is used to analyze the queuing performance of generic variable length packet networks for the first time.The self-similar traffic is generated by multiplexing a large set of independent Pareto heavy-tailed interarrival ON/OFF sources.The simulation results show that the heavy-tailed traffic results in queuing performance deterioration for variable length packet networks,which is in accordance with the analytical results for ATM switches.