IP流到达分布研究

Study on IP flows arrival distribution

在探讨现有IP流到达分布研究成果的基础上,首先对不同来源的大规模网络的IP流到达时间间隔分布和自相关性进行了分析,发现IP流到达时间间隔均服从Weibull分布,但随着IP流到达密度的增加,Weibull分布的参数α逐渐接近于1,则Weibull分布退化为指数分布;IP流的自相关系数随着被考察IP流流长的增加逐渐减弱,流长为100左右的IP流,其到达过程彼此基本不相关。

Based on the former research on IP flows arrival distribution, this paper analyzed the interarrival time distribution and the autocorrelation of IP flows coming from different large-scale networks. It is found that the IP flows＇ interarrival time distribution follows Weibull distribution whose parameter α is close to 1. 0 as the scale of network becomes larger, so the Weibull distributions degrades to exponential distribution. And the autoeorrelation parameter becomes less as the IP flows＇ length increases---when the flow lengtb of a IP flow increases to 100,its arrival becomes irrelevant to others.