Research Of Rate Mechanism On Internet Congestion Control Protocol

With the development of network and communication technology, the Internet has become an indispensable part of people’s daily lives. Although the Internet bandwidth is increasing with the development of network hardware, the network resources are limited. Because of the unbalanced distribution of network resources and network traffic, network congestion is the inherent nature of the Internet. In order to guarantee stability of the Internet, it is necessary to execute congestion control. According to the position of the congestion control execution, congestion control algorithm can be roughly divided into two categories, source algorithm and link algorithm. Source algorithms, such as TCP, are often executed on the edge of the host and network equipment, and adjust the sending rate according to the host information. Link algorithms, such as explicit congestion control algorithm, are used to collect information from network devices of congestion, detect the occurrence of congestion and feedback the congestion information to the host through the packet head information.The traditional congestion control mechanisms always depend on TCP or improved TCP protocols based on end-to-end source algorithms. With the development of Internet, the limitations of traditional congestion control mechanisms expose. This thesis analyzes the basic principle of TCP congestion control mechanism, including the slow start, congestion avoidance, fast response and fast retransmission mechanism, and some improved TCP algorithm, such as TCP Westwood, TCP Vegas, HSTCP. The congestion window mechanism of TCP discomfort streaming media applications and the high speed Internet, and it cannot make full use of bandwidth in wireless networks because of channel errors. To solve these problems, this thesis carries out a RCP(Rate Control Protocol) based exact rate feedback congestion control algorithm through the research of the explicit congestion control algorithm, and put forward a bandwidth estimation based wireless network rate feedback congestion control scheme, finally propose a kind of rate based of multipath transmission control scheme. The concrete research content is as follows.(1)There are some limitations exist in the method of flow amounts estimation through deep analysis for RCP protocol. Aiming at this problem, this thesis propose a RCP based exact rate feedback congestion control algorithm which can be used to accurately calculate the feedback rate without estimate the amounts of flow through a router. Through various simulation experiments on the NS2 platform, the results show that the algorithm can achieve high network utilization as RCP, and it has lower average queue length and shorter flow completion time. This algorithm can improve the Qo S of network.(2)Through analyzing the performance and limitations of TCP congestion control in wireless networks, the throughput will fell sharply in the wireless environment with high error rates. The main reason is that the existing TCP congestion control mechanism cannot know whether the wireless link causes packet loss, while TCP still reduce the transmission rate. The existing bandwidth estimate congestion control mechanisms based on end-to-end also have the TCP defects, such as the performance is not well in high bandwidth wireless links, throughput jitter is excessive when packet loss and utilization rate is low. This thesis proposes the bandwidth estimation based rate feedback congestion control algorithm in wireless networks. According to the NS2 simulation analysis, it can achieve higher throughput and better stability.(3)How to ensure the reliable transmission of multiple paths is very important in end-to-end multipath transmission of multipath transmission control protocol. This thesis puts forward a rate based multipath transmission control schemes, using the accurate rate feedback congestion control algorithm in every subflow of MPTCP, fully improving the throughput of each sub-flow. It can quickly use residual bandwidth and has faster convergence speed and higher throughput through theoretical analysis.