IP数据网络混和流的特性分析及模型化

Characteritics Analysis and Modeling of IP Data Network Aggregate Traffics

IP数据网络流量分析模型的研究一直是通信网络性能分析中一个及其重要的问题。许多文献提出一些不同的流量模型,其中包括;马尔可夫模型、回归模型、长程依赖流量模型和(σ,ρ)漏桶模型。这些模型描述了网络流量的一些特性,并具有特定的使用范围。但是随着网络技术的发展和新网络应用的不断涌现,网络流量呈现出一些新的特性。通过对实际网络测试结果的时频分析,发现网络混合流的包速率变化呈现出一定的周期性,而且具有非平稳正态特性。基于此结果,该文提出了网络混合流的周期非平稳正态模型。

One of the focuses of any performance evaluation of IP data networks is their analytical Modeling of traffic source.Many papers proposed the different traffic models,which included Markov Models,Regression Models,Long-Range Dependent Traffic Models and(σ,ρ)Leaky-Bucket Model.These models described some characteristics of network traffics,and had the specific applications.With the development of network technology and the emergence of new network services,however,the network traffics present some novel characteristics.Through the time -frequency analysis of the measuring results of the factual network,it is found out that the packets rate of aggregate traffics exhibits the periodicity,and has the characteristics of non-stationary norm processes.This paper proposes the Periodic Non-Stationary Normal Model of aggregate traffics based on measuring and analyzing results.