Simulation Based Analysis of the Performance of Self-similar Traffic
There are large number of experimental evidences that network traffic processes exhibit ubiquitous properties of selfsimilarity and long-range dependence (LRD), i.e., of correlations over a wide range of time scales. Modeling and performance analysis of self-similar traffic have become an investigating hot topic in computer network. However, most of the studies have been focused on the estimation and influence of Hurst index, and ignored the other factors. In fact, some other factors have also important influence on network performance. In this paper, we make a thorough investigation on the influence factors to the network performance of self-similar traffic. Based on the buffer overflow probability derived by Norros, we firstly derive the formulas of average queuing length, queuing length variance, average delay, jitter and effective bandwidth. Then the influence of Hurst index, average arrival rate and variance coefficient of the traffic, along with the buffer size, utilization and the effective bandwidth of the system on the performances of self-similar traffic are investigated by means of simulation. The results reveal that all these factors have great influence on performance of the self-similar traffic and there is evidently time scale effect among them. Finally, the critical time scale is derived.
Self-Similarity Long-Range Dependence Traffic Modeling and Performance Evaluation Fractional Brownian Motion Time Scale
TAN Xianhai ZHUO Yiwen
School of Information Science and Technology Southwest Jiaotong University Chengdu 610031,China School of Civil Engineering Southwest Jiaotong University Chengdu 610031,China
国际会议
第四届国际计算机新科技与教育学术会议(2009 4th International Conference on Computer Science & Education)
南京
英文
312-316
2009-07-25（万方平台首次上网日期，不代表论文的发表时间）