A Study Of Congestion Control Strategies On The TCP-friendly

These two decades have seen a boom in the use of TCP/IP-based Internet. The size of the network is growing larger; topological structure is becoming more complex; the number of users is increasing dramatically; and the application of the net is becoming increasingly diversified. However, consequently the congestion problems are getting more and more serious with each passing day. Congestion can give rise to a sharp drop in such network performance metrics as packet loss and throughput, which in turn will greatly influence the unilization of network resources. Therefore, congestion control has been a hot issue.Internet relies mainly on the TCP end-to-end congestion control to avoid network congestion, but this strategy can not satisfy the requirements of the various applications in the complex network. The appropriate introduction of congestion control mechanisms in the middle nodes can help detect congestion much earlier and thus can prevent or at least reduce congestion problems. Hence, router congestion control has aroused the concern of many researchers recently. With the appearance of audio, video and other real-time multimedia applications, there is an increase of UDP traffic on the Internet. Since UDP flow does not contain a congestion control mechanism, the problem of TCP unfriendly flow comes into being. TCP flows cannot get fair band width, even "starves", which will lead to the collapse of the network at worst and seriously affect network performance. So there is a need for congestion control for UDP flows to ensure the friendliness of TCP flows.This thesis discusses the current congestion control algorithms from two perspectives:the source end system and the router. Based on this is an in-depth study of the TCP-friendly congestion control. The main contents of the thesis are as follows:Firstly, an active queue management algorithm NF-RED is proposed in view of the simultaneous transmission of UDP flows and TCP flows on the Internet. This algorithm identifies the non-TCP friendly flows according to the historical packet loss information and by setting a threshold, and achieves the fairness of their coexistence through the reinforcement of punishmentSecondly, this thesis puts forward a TCP-friendly congestion control mechanism EFCC based on the source end and the router. Through the cooperation of the terminal host and the routers, the sending rate of UDP adjusts according to the feedback from the receiver and the congestion is supposed to be detected early with the help of the router.Finally, this study provides the experimental verification of the proposed algorithm. With the co-transmission of UDP and TCP, the proposed algorithm is much better than the original one in terms of throughput and fairness, etc.