TCP网络拥塞控制算法比较研究

Comparing Study of TCP Congestion Control Algorithms

网络应用的快速发展使得网络拥塞问题变得越来越严重,因此,拥塞控制机制的研究已成为网络鲁棒性研究热点之一。分析了4种TCP拥塞控制典型算法(Newreno、Vegas、Sack1、Fack),并在NS仿真环境下进行了仿真,着重分析了4种算法的拥塞窗口、吞吐量、丢失率和平均队列的变化情况,总结了4种算法的优缺点。仿真结果表明,Vegas算法采用时间延迟作为拥塞度量,通过比较实际传输速率和期望传输速率之间的差值来预知拥塞,能较好地预测网络带宽使用情况,其数据包丢失率小、缓存队列长度较小,具有良好的应用前景。

With the rapidly improvement of computer network,congestion control became serious gradually.Therefor,congestion control techniques became the hot issue and difficult issue of the Internet research on robustness.This paper introduced the TCP congestion control algorithms such as Newreno,Vegas,Sack1 and Fack in brief,and analyzed their characteristics.The NS simulator was used to evaluate their performance about congestion window,throughput,loss rate and the change of average queue.By analyzing and comparing,the simulation results show that Vegas algorithms for delaying measure of the congestion,meanwhile calculate the network congestion by the difference between actual transfer and expected speed,it is used to forecast the utility of network band well,the package loss rate is less than other three algorithms,Consequently it declines the length of buffer queue,and has a good application prospect.