网络流量分布特性是众多网络应用的基础。提出一种网络流量特性的综合分析方法。首先对获取的流量样本进行时间尺度化处理,之后引入半马尔可夫模型描述网络流量特性,通过忙阈值和闲阈值的设定将网络流量划分为三种状态:忙、中间、空闲,研究各状态下的网络流量特性及各状态间的相互转换关系,然后建立了基于参数估计和非参数估计分布模型库,最后进行分布拟合。该方法适用于对所有协议的网络流量分析,比使用单一分布模型分析更合理。实例分析证明该方法是正确、有效的。 Network traffic distribution is the basis of many network applications. A comprehensive analysis method of network traffic characteristics is proposed. First of all, the obtained traffic samples are time-scaled, and then the semi-Markov model is introduced to describe the network traffic characteristics. The network traffic is divided into three states by setting the busy threshold and idle threshold: busy, intermediate, and idle State network traffic characteristics and the mutual conversion between the state, and then set up based on parameter estimation and non-parametric estimation distribution model library, and finally distribution fitting. This method is suitable for network traffic analysis of all protocols and is more reasonable than using single distribution model analysis. The case study proves that the method is correct and effective.