自相似业务模型下的队列分析──大偏差技术

The Queueing Analysis of Self-Similar Traffic Models Large Deviation Technique

近年来对许多实际运行中的网络（局域网和广域网）的测量与分析证实：真实的网络业务具有统计上的自相似性而且是长相关的。传统的排队理论均假定业务是Poisson或Markovian，因此是短相关的，那里的队列分析方法对自相似业务模型是不适用的。大偏差技术是对自相似业务模型进行队列分析的有效方法。本文应用大偏差技术对以FARIMA为业务模型的队列系统进行了分析，结果证明：队列长度尾分布渐进为Weibullian分布，而且与模型的短相关结构无关。

A number of recent measurements and studies of actual traffic from working networks (LAN and WAN)demonstrated that real traffic has statistical self-similarity and,therefore,is long range-dependent. In the traditional queueing theory the traffic is assumed as Poisson or Markovian, so is short-range dependent. The methods there are basically not applicable to self-similar traffic. The large deviation technique is an efficient approach to queueing analysis with self-similar traffic. In this paper we apply large deviation technique to queueing analysis with FARIMA traffic and show that the tail distribution of queue length is asymptotically Weibullian and is independent of the short-range dependent structure of the model.