多重分形网络流量的单队列输入排队性能研究  

Single Queuing Performance Study with Multifractal Network Traffic Input

作　　者：魏进武[1] 汪斌强[1] 邬江兴[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,河南郑州450002

出　　处：《电子学报》2006年第4期658-663,共6页Acta Electronica Sinica

基　　金：国家"十五"863计划信息技术领域重大专项(No.2003AA103510No.2004AA103130No.2005AA121210)

摘　　要：多重分形已被证明能够较好地用于网络流量多尺度突发行为特性的研究.本文以队列处于满负荷状态的概率为指标研究了多重分形流量输入下单队列的排队性能问题,提出了以尺度函数与矩因子表示的队列满负荷运行概率的解析性框架.用FBM过程作为输入,从理论上验证了该概率框架的正确性,进一步用实测的网络流量数据对多重分形流量输入下单队列满负荷运行的概率进行了分析与比较,结果表明该框架能同时满足单、多重分形流量输入时排队性能的计算.结合文献[1]给出的联合多重分形(JMF)流量模型,为排队系统中队列深度的设计提供了有效的定量计算方法.Multifraction has been proven suitable for characterizing the burstiness of network traffic in multi-timescale. This paper studies the queuing performance of single queue driven by a general multifractal input process using the queue over-loading probability. We proposed an analytic probability framework to describe the queue over-loading probabilities constituted by scaling function and moment factor, which is validated by the analytical results under the conditions of FBM process inputting. Furthermore, the probability performances are comprehensively analyzed and compared as queue over-loading with an actual traffic data inputting. The results indicate that the analytic probability framework is available to calculate the queuing performance driven by both momfratal and multifractal input processes. Combining with the joint multifractal （JMF） traffic model presented in literature[1] ,the queue length of the queuing system can be calculated quantitatively by the probability framework.

关 键 词：因特网 网络流量 排队性能 多重分形 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]