网络性能评价中重要事件采样仿真

SAMPLING OF IMPORTANT EVENTS IN PERFORMANCE EVALUATION OF NETWORKS

介绍了重要事件采样仿真技术 ,分析了其仿真原理和实现方法 ,给出了典型的仿真算法 .仿真表明 ,对于分数整合滑动平均过程为输入的单服务队列 ,系统的溢出率与业务的长相关有密切关系 .仿真结果与大偏差等分析结论较为一致 ,说明了仿真的可靠性 .

This paper proposes the sampling technology of important event in performance evaluation of networks.The theory and typical algorithm are discussed.Given a queue fed by a FARIMA process,the simulation results show that the long range dependence of traffic gives rise to the overflow of queue.Both the decrease of the time of simulation and the simulation results are in accordance with the analysis result of large deviation technology.