摘要
网络业务建模是网络规划与性能评价的重要基础,常用分形布朗运动(FBM)模型来进行自相似业务建模。但FBM是精确自相似过程,由于比较简单,它只能用于刻画单一Hurst系数的精确自相似过程。而实际网络流量是一个复杂的过程,其统计特性随时间尺度而变化,不能用单一Hurst系数的精确自相似过程来描述。本文提出用多个FBM模型对尺度行为建模,并在Norros给出的缓冲区溢出概率公式的基础上,推导出平均队列长度、队列长度的方差、平均时延、时延抖动和有效带宽的计算公式。
Traffic modeling was an important basis of network programming and performance evaluation. FBM was only efficient for approximating the performance of a class of exactly self-similar traffic, whose correlation property could be described by a single Hurst index. However, most traffic in real networks did not display exactly self-similarity, but had more general long-range dependent properties, which could not be described by a single Hurst index. By paying attention to the time scales dominating the performance, in this paper, it was proposed an (MK)-FBM method to solve the problem. Based on the buffer overflow rate given by Norms, it was derived the formulas of average queuing length, queuing length variance, average delay, delay jitter and effective bandwidth.
出处
《铁路计算机应用》
2009年第6期10-13,共4页
Railway Computer Application
基金
国家自然科学资金项目(60572143)
西南交通大学科学研究基金项目(2005A03)
关键词
业务建模
自相似
分形布朗运动
Hurst系数
traffic modeling
self-similarity
Fractional Brownian Motion(FBM)
Hurst inde