Research For Active Queue Management Algorithm On Self-Similar Traffic Network

With the development of network technology the network scope extended rapidly, especially in 1990s', the Internet based on IP presented a tremendous increase and Internet has grown up to be the essential global information facility. With the network of new-pattern avalanched and the network user increased rapidly, the Internet data flow is increasing sharply. Internet is no longer a network only for transmitting data; it has turned to be the most important method of information communication and the most compositive network which carry all kinds of multimedia information, such as data, sound, video etc.In recent years, a great deal of examining and analyzing on actual network traffic indicated it has distinct self-similarity,which breaks the old basic hypothesis that network flow is short range dependence. The traditional Poisson model is no longer appropriate. Due to the great differences between the network capability of self similar and that of traditional model, the network management, congestion control and protocol design of the two models are also different.The rapid development of network and the implement of all kinds of network business make the Internet a huge and intricate network. As a result the problem of network congestion has come forth inevitably, and it causes a decreasing business index and a inefficient network. So people are paying more attention to the congestion control which is going to be a important rule to guarantee the robustness and network working smoothly.In congestion control, Active Queue Management algorithm working on network node is a important and effective way to solve the network congestion and guarantee QoS. So research on this field is definitely significant.This dissertation concentrated on the management of buffer and improved the classic Random Early Detection after analyzing the influence of self-similar on network capability. Added the function of self-adaptive, make the network adjust its own parameter to adapt itself to the actual network traffic. The main research results are presented as follows.Did research on Random Early Detection and other AQM algorithm, analysed their advantage and shortcoming.Started with the phenomena of the self-similar, studied the definition, characteristics, estimate, model,forecast of self-similarity and reason of its coming out. Afterward, programming to implement the ON/OFF model based on the Pareto chorology. This model simulated the real network environment and had network traffic with the characteristics of self-similar.Analysed the influence of self-similar on the network capability, and improved the old algorithm. Using the AQM evaluating standard and proposed the new Based Interval Random Early Detection which is suitable for self-similar network traffic model. Furthermore, implemented and simulated the algorithm. The simulating result showed the new algorithm is better than RED on reducing the system burden, stabilizing the average queue length and improving the transient performance of the system.On the basis of Based Interval RED, the algorithm was improved and added the function of self-adaptive. It can adjust it own parameter according the network traffic to adapt itself to the real network environment.