The Study Of TCP Congestion Control In High Speed Networks

As we know, TCP is the most important transport layer protocol and the main part of the end-to-end congestion control in Internet, its performance in different network environments always receives wide attentions. The performance of the TCP is mainly determined by the congestion control algorithm. The high-speed network with a large delay-bandwidth product is gradually appearing in the actual production and living environment as the network technology development at the same time. In the recently years, a number of approaches have been proposed to improve the TCP performance. This paper studies on the TCP congestion control algorithm in high-speed network and mainly includes the following three parts of the work:proposing a new TCP congestion control algorithm、using the NS-2simulation platform to verify the performance of the new algorithm and implementing it as a kernel module on Linux system and then accomplishing some test in the real network.Today there are many different types of high speed TCP congestion control algorithm in the network. They can be divided into loss-based, delay-based and the mix of the two before. But the loss-based protocols may severely decrease the throughput of regular TCP Reno flows which are widely used in the network. On the other hand, pure delay-based approaches may not work well if they compete with loss-based flows. In this paper, we propose a new hybrid control algorithm of TCP congestion control algorithm called Hybrid High Speed TCP (HHS-TCP), which is a synergy of delay-based and loss-based approach. The packet sending rate is determined by the two mechanisms together. The new algorithm can be aggressive and increase the sending rate when the link have not been fully utilized but gracefully retreat in a busy network when a bottle-neck queue is built. The design of HHS-TCP is to satisfy to efficiency requirement and TCP friendliness requirement simultaneously.In order to verify the performance of the new algorithm, we implement a comprehensive test including effectiveness and fairness testing using the NS-2simulation platform. Through the measurement of the throughput of a single flow in the random loss link and the measurement of the throughput of multiple streams under two different routing strategies which are the Drop-tail and the RED mode respectively. The result indicates that the HHS-TCP can use the network bandwidth more efficient. At the same time, by simultaneously comparing the throughput ratio between a certain type of high-speed protocol and Reno in a same bottle link. As we can see that the HHS-TCP with a hybrid mechanism can obtain the ideal fairness. On the other hand, in order to test the performance of the algorithm in a real network and based on the analysis of the Linux TCP protocol stack structure, we implement the new algorithm in the Linux system in a module form and it’s easy to be loaded and removed. We do the test in a LAN by measuring the speed of the file transmission. The result validates the performance of the new algorithm can improve transmission speed by10%-15%.The main contribution and innovation of this paper are:By studying the problem of the congestion in the network and the TCP congestion control algorithm, we analyze the cause of the congestion and the evaluation method of the congestion control algorithm and then we propose a new hybrid mechanism based congestion control algorithm which meets the high bandwidth utilization and fairness to the traditional protocol. We implement the algorithm as a kernel module in the Linux system. Actual tests show that the new algorithm has some practical value.