Research On Characteristics Of Self-Similar Traffic In Transmition Propocess

The characteristics of network traffic have great changes with the rapid development of computer network technology. Over the past decade, a large number of study results show that network traffic exhibits ubiquitous properties of self-similarity and long-range dependence(LRD), which have great influence on network traffic modeling, performance evaluation, admission control and so on. So self-similarity of network traffic has become one of main factors needed to be considered while making network planning and design.In this thesis, the changes of self-similarity of network traffic and the factors impacting it are studied in detail. Firstly, mathematical definition of self-similarity is given. The properties of self-similarity and the methods of estimating H parameter are depicted. Then the aggregation charateristic of self-similar traffic and poisson traffic, the impact of traffic shaping mechanism on self-similar traffic and the characteristic changes of self-similar traffic produced after passing through the network switch node are studied.On the aspect of traffic aggregating, the traffic properties change after short-range dependent traffic and long-range dependent traffic be aggregated is analyzed. Traffic model is important to network simulation, Poisson and FBM are used as short-range dependent traffic model and long-range dependent traffic model respectively. and presented them in detail before using them.Performance of the aggregating traffic, such as the changes of H parameter, average queuing delay and packet loss rate, are analyzed.On the aspect of the impact of traffic shaping on the characteristic of self-similar traffic, token bucket is used to shape the self-similar traffic. The systhetic traffic based on FGN are passed through the token buchet and the property change is analyzed, focusing on the change of Hurst parameter and variance of traffic.In the same time, we have also detailedly studied other parameters having effect on performance of network.Network traffic is dynamic and it will pass through network switching node while travelling from source to destination. Few studies were performed so far and their study results show that the properties of self-similarity of traffic will change while travelling through the network switching nodes. Some problems are still needed to make further studies. In the thesis, we have studied these problems deeply.