摘要： TCP Reno拥塞控制机制是目前互联网中采用的主流拥塞控制算法。根据TCP Reno实现拥塞避免与拥塞控制的AIMD算法中加性因子与减性因子过于武断，对可用带宽的探测缺乏细分，造成信道利用率未达合理水平等缺陷，为此，提出一种根据拥塞窗口的增长情况进行更为平滑的信道容量探测算法，采用基于对数的增长算法和下降算法，通过NS仿真对吞吐量、公平性和友好性进行评估。仿真结果表明，该改进TCP拥塞控制算法吞吐量比TCP Reno提升10%。

关键词: 传输控制协议, 网络拥塞, 拥塞控制, 加性增长乘性减少算法

Abstract: TCP Reno congestion control mechanism based on the algorithm of Additive Increase Multiplicative Decrease(AIMD) is mainly used in the Internet in the present. Aiming at the drawback of this algorithm in dealing with the detection of available bandwidth utilization rate due to AIMD factors too arbitrary, traditional AIMD algorithm is researched and a smooth channel capacity detection algorithm based on the growth of a congestion window is proposed, its throughput, fairness and friendliness are evaluated by NS simulation. Simulation results show that the throughput of modification algorithm of TCP outperforms TCP Reno by 10%.

Key words: Transmission Control Protocol(TCP), network congestion, congestion control, Additive Increase Multiplicative Decrease(AIMD) algorithm

中图分类号: 

• TP393