基于最小均方滤波的RTO预测算法

传统的传输控制协议(TCP)中的重传超时值(RTO)设置过于保守,且RTO预测算法对于往返时延(RTT)变化的响应过于缓慢,造成长时间的无谓等待,降低了带宽利用率。为了解决这些问题,提出一种基于最小均方(LMS)滤波的预测算法。为了验证新算法的性能,选择了大量有代表性的网址进行试验,对新旧2种算法的预测结果进行比较,并做了性能分析。试验结果证明,所提出的算法相比于原算法,具有能很快达到稳定,稳定后的预测误差小,对RTT变化的响应快等特点,提高了传输性能。

RTO prediction algorithm based on a least mean square filter

The retransmission timeout(RTO) prediction algorithm that current the transmission control protocol uses is too sluggish to respond to changes in the round-trip time(RTT).Furthermore,the predicted values are usually much larger than real RTTs,leading to reductions in the bandwidth utility ratio.A prediction algorithm based on a least mean squares filter was developed to predict better values.The algorithm performance was evaluated using a variety of typical websites.The results show that the algorithm quickly stabilizes,has a very small prediction error,and quickly responds to RTT changes,so it can improve transmission performance.