Research On Congestion Control Mechanism In Heterogeneous Networks Based On RTT

With the current explosive growth of social information, Internet as an important medium of information dissemination in the past few decades have also developed rapidly, accompanied by congestion problems are more serious. The direct result of the network congestion is the performance decline of the whole network which includes:the packet loss rate increases, the end-to-end delay increases, the network throughput decreases, and the network system may go through congestion collapse. Developed an effective congestion control techniques has been imperative. In general, the root causes of network congestion are that the network load to be generated by client systems is greater than the capacity of network resources and processing power. When the network is in the state of congestion collapse, the network throughput will decreases sharply with tiny load increment.Congestion control technology since 1988, Van Jacobson was first raised in their papers so far, has gone through 22 years of development history. From RFC 1122 to TCP Tahoe and to the most widely used TCP Reno, each modification has made a big enhancement in its performance. Congestion control is the basic part of QoS, and the precondition of basic QoS is the use of congestion control. Therefore, it is very important to solve the problem of congestion of networks. Since TCP Reno is currently the most widely used transport layer protocols, so this paper will focus on TCP congestion control algorithms.TCP Vegas is one of the most promising mechanisms that may substitute TCP Reno, which is most widely implemented in the current Internet. However, the weakness that TCP Vegas shows in competing with TCP Reno over mixed network scenario has prevented it from being widely implemented. The weakness of TCP Vegas in mixed network scenarios is analyzed, and a new slow start algorithm is presented. The analytic result shows that the improved slow start algorithm can generate a better initial value for congestion avoidance, so that the throughput is improved.By establishing its mathematical model, we confirm the validation of the modification.This paper first introduces the current research broadband network congestion situation, and summarized researches of TCP congestion control mechanism and pays attention to improving for TCP Vegas.The major contributions of this thesis are as follows:1. Focusing on the phenomenon of TCP Vegas to change from slow-start phase to congestion avoidance phase too early, and the problems of congestion window is too small resulting in bandwidth utilization is dropping, and so on with Reno congestion based on packet loss to determine when the algorithm competitive bandwidth resulting in fairness badly.For Vegas-A algorithm the shortcomings that the proposed improved algorithm Vegas-O. When in slow-start phase, with the range between the index and the linear growth rate to increase the congestion window size, so that it can quickly achieve the value of available bandwidth and reduce the incidence of conflict; The congestion avoidance strategy is based on Vegas-A for improvement. In addition, within each RTT the incremental value of the congestion window size also made the appropriate changes, it can be a better solution to the slow start-up phase of their problems. Vegas-A algorithm is better than the increase in the active window, but also in the improved stability while retaining the advantages of the original algorithm, making the algorithm more adaptable than the original algorithm for the network conditions change. When the co-existence with the Reno, this algorithm is compared with Vegas-A algorithm performance is good, especially in high bandwidth-delay product networks.2.In wireless network,channel interference and noise cause significant performance degradation of TCP. In order to solve the problems existing at the application of TCP Vegas,based on TCP New Vegas algorithm,combining the advantages of TCP Vegas A+ algorithm,an improved algorithm termed TCP Vegas-P was proposed.A comprehensive improvement was taken to the premature end of the slow start phase and congestion avoidance in the reverse link congestion resulting in the decline in throughput.The new algorithm on the slow start and congestion avoidance phases corresponding improvement, and finally to combine. By the simulation results show that improved algorithm in the wireless network to be able to better estimate RTT, it has a good performance to the mentioned problems about Vegas, and can be well applied to wireless network environment.