Research Of End-to-End TCP Congestion Control Algorithm

Network congestion is the reduced network efficiency due to the limited storage and limited processing speed of routers or the ends failing in using the limited network resources properly.According to the statistics,more than 90% of the traffic on the Internet is carried by the Transmission Control Protocol(TCP).It is very important to design a high-efficient TCP congestion control algorithm for the development of the computer network.First,to improve the performance of TCP CUBIC congestion control algorithm,a new delay-based TCP congestion control algorithm named TCP CUBIC-R was presented.TCP CUBIC-R uses round trip time(RTT)to determine the degree of network congestion to modify the original congestion window size calculation function of TCP CUBIC algorithm.Experiments reveal that proposed algorithm has a 5% to 7% of improvement in performance,compared with the TCP CUBIC congestion control algorithm.Then,focus on the problems that the traditional congestion control algorithms can not adapt to various networks fully and cat not set congestion window size flexibly,this paper presents a new TCP congestion control algorithm based on network status classification(CCC).This algorithm can generate congestion control strategies that suitable for a specific network automatically.Furthermore,the congestion control strategies can be adjusted online in response to the changes of the network.Experiments show that the proposed algorithm achieves better network throughtput and fairness than the traditional congestion control algorithm.