Research On The Congestion Control Algorithm With Self-Similar Traffic

With the rapid development of society, the thransmission of information is more and more quickly and much more than ever. The amount of information on the Internet is exploding, so the network bandwidth and hardware resources such as cache capacity can not meet the growing demand for business flows, resulting in network congestion. Network congestion decreases the stability of congestion window size and instantaneous length and the system performance such as response speed severely, resulting in the poor network quality of service. Therefore, congestion control has become research focus in the areas of network communications in recent years. On the other hand, a large number of studies show that traffic of the current network has universal self-similarity and long-range dependence. Which makes protocol,strategies and evaluation methods on the network based on original Markov model is not accurate enough, resulting in increment loss rate and the degradation of network performance. On the other hand, as self-similar flow is predictable, it also provides a new way to congestion control algorithm.In this thesis, by considering the self-similar characteristic, a new method of self-similar congestion control is discussed, and the performance of traditional congestion alogrithm is compared with self-similar congestion control algorithm through simulation. The thesis focuses on application of HSTCP congestion control algorithm in the networks under self-similar and uses self-similar characteristics to appropriate modify the formula of RTT in order to improve the network performance under the environment of self-similar traffic. The main research works and achievements of this thesis are as follows:Firstly, the common congestion algorithms is studied, and the basic principle and merits and faults of TCP Reno, Vegas algorithm are analyzed. The performance of these algorithms with traditional short-rang dependent traffic input is compared with that of with self-similar traffic input by simulating with OPNET Modeler 10.0.Secondly, the methods of the traffic prediction are studied, and the feasibility of traffic prediction is confirmed in theory.Thirdly, another improved HSTCP congestion control algorithm is proposed under the environment of self-similar traffic input. The predictability of self-similar flow is used to forecast the congestion window size of the next time. Then the forecast window is applied to HSTCP congestion control algorithm to dynamically adjust the window parameters. It is able to reflect the self-similar network environment better than the method of traditional. The effectiveness of the improved algorithm is veritied through simulation. Finally, because the characteristic of self-similar must be reflecting to RTT, in this thesis we use this feature to modify the RTT valuation formula. It can be more suitable for environment of self-similar network. The new RTT value is used to detect the grade of congestion and adjusted dynamically the congestion window based on the congestion levels. The effectiveness of the improved algorithm considering the change of the RTT valuation formula is verified through simulation.