自相似业务流下的排队性能分析

Analysis of Queuing Performance of Self-Similar Traffic Input

根据到达时间间隔服从Pareto分布的业务具有自相似性这一结论,利用剩余累积分布函数拟合和矩阵几何方法研究了Pareto分布到达时间间隔和负指数分布服务时间下的网络系统排队性能,并获得了Pareto/M/1/K队列平均排队队长和缓存溢出概率的近似结果.仿真结果表明这种方法具有较高的精确性.

As all the traffics whose arrival time interval submits to Pareto distribution have the self-similarity, the queuing behavior of the network system with Pareto distribution inter-arrival time and negative exponential distribution service time distribution is investigated via the fitting technique of complementary cumulative distribution function and the matrix geometric method. The average queue length and buffer overflow probabihty of Pareto/M/1/K queue are then approximately derived. Simulated results are finally presented to verify the accuracy of the proposed method.