Congestion Control Algorithms In Transport Layer Of Computer Network Based On Self-similar Model

With the development of computer science and internet, people are not only focusing the interests on traditional data transfer in network but also multimedia traffics such as audio and video. At the same time, it may also bring some troubles. For instance, a new property of network traffic (self-similarity) has been observed in diverse networking contexts. Some old protocols, policies and evaluate methods which are based on previous Markov network model are not suitable for this self-similar traffic. And when TCP data share bandwidth with UDP data that is popular for multimedia applications over the internet, it also causes fairness problem. However, even recently, these are not many excellent solutions. It is because that network designers have left few control interfaces in the lower layer of the network, and network users have strong unwillingness to make major changes on the accustomed protocols. In the upper layer of the network, there do exist some control methods, but most of them only forcus on the QoS of a single connectionand make very little contribution to the total network. In this paper, based on analysis of network protocols and property of self-similarity in the network traffic, a class of new algorithms of congestion control is presented and theoretically analyzed. The new algorithms need not to add any unacceptable matters in old protocol, it only need a few changes on the codes at the traffic sender. Simulation results show that network performance is improved in different long-range dependent network circumstances and not worse than the old algorithm when applied in short-range dependent network environment.