基于峰值流量的网络行为特征及影响因子分析

Analysis of network behavior characteristic and influence factor based on the peak traffic

基于网络流量的突发性提出峰值流量测度,建立了一种评价网络运行状况和规划容量的方法。利用统计学方法(方差分析和协方差分析)对21个CERNET校园网的内在特征和峰值流量进行实验分析,发现峰值流量间相互独立且服从高斯分布,网络内在特征对峰值流量的影响存在显著差异,以及链路带宽需求与网络用户数之间存在较强相关性。由此建立线性回归模型及容量规划模型。实验表明建立的容量规划模型能够对新建校园网的接入带宽进行准确评估。

Based on the burstiness of network traffic, the peak traffic metric was introduced, and the method which was used to evaluate network operation and plan capacity was proposed. The network intrinsic features and the peak traffic for the 21 CERNET campus networks were analyzed by statistical methodologies （analysis of variance and analysis of covariance）, so that the peak traffic： is mutually independent and follows Gaussian distribution, the network intrinsic features have dominant impact on the peak traffic, and the number of network users is highly related to link bandwidth demand. On the basis, the linear regression model and capacity planning model were constructed. Experimental results illustrate that the proposed ca- pacity planning model is used to evaluate the access bandwidth of new campus networks exactly.