突发信源的排队分析  

Study of the queueing of bursty traffic sources

作　　者：周亚建[1] 李建东[1] 

机构地区：[1]西安电子科技大学信息科学研究所,陕西西安710071

出　　处：《西安电子科技大学学报》2002年第1期1-5,共5页Journal of Xidian University

基　　金：国家自然科学基金资助项目 (69872 0 2 8);国家"863"计划 (863 3 17 0 3 0 3 99);高等学校骨干教师资助计划

摘　　要：提出了利用相位法进行突发信源排队分析的方法 .突发信源是ON/OFF信源的一种 ,其ON期和OFF期分别服从独立的重尾分布 (如Weibull分布 ) .如果把ON期内产生的所有分组作为一个批量 ,那么OFF期对应着批量的到达过程 ,而ON期对应着批量的服务过程 .利用相位法 ,ON期和OFF期的长度都可以用多个服从指数分布的随机变量的和来精确地逼近 ,即批量的到达和服务过程都可以分成多步完成 ,而且每一步的长度都服从指数分布 .这样就可以用Markov链对突发信源的排队性能进行分析 .仿真结果表明 ,只要合理地选择相位逼近的参数 。An approach, derived from the method of phases, to analyzing the queueing performance of bursty traffic sources is presented. A bursty traffic source is an ON/OFF process, where ON and OFF time length distributions are heavy tailed (e.g., Weibull distributed). If all the packets arriving during one ON period are considered to form a batch, OFF periods correspond to the arrival intervals of consecutive batches, and ON periods correspond to the service process. According to the method of phases, both ON periods and OFF periods can be approximated by the sum of several exponentially distributed random variables. Based on the assumptions above, arrival intervals and service intervals can be divided into several steps and the length of each step is exponentially distributed. Therefore, the Markov chain can be employed here to analyze the queueing performance of bursty traffic sources. Simulation results show that satisfactory result can be obtained with appropriately selected parameters.

关 键 词：相位法 突发信源 WEIBULL分布 排队分析 

分 类 号：TN911.21[电子电信—通信与信息系统]