基于流量预测的TCP慢启动算法

Study of TCP Slow Startup Algorithm Based on Flow Farecasting

分析了标准慢启动策略,针对现有策略中拥塞窗口指数方式增长不够平滑的问题,提出一种基于往返时延(RTT)流量预测的TCP慢启动改进算法(flow-forecast-Based Slow Start algorithm,FFB-SS)。在慢启动后期引入基于RTT控制因子,使窗口增加缓慢。用NS2进行仿真分析的结果表明,改进算法可以在慢启动后期减缓拥塞窗口的增长速度,丢包率明显减小,瓶颈链路的利用率也有所提高。

This analyses the standard slow-start strategy,and proposes a new method of slow-start strategy based on flow forecasting,the problem that the congestion window grows unsmoothly is solved. At the stage,window control growth factor is used based on Round-Trip Time(RTT) to slow the increases speed.A simulation is employed with NS2 to compare the performance of the present ed strategy and the new strategy. The result shows that the new method adaptes the congestion window grows more smoothly,the package loss rate decreases,bottle-neck link usage rate increases.