一种快速的报文丢失率推测方法

Fast approach to inferring loss performance

目前,在网络断层扫描的推测阶段主要采用的方法是似然估计,这些方法的计算量会随着网络规模的增加而急剧增长,从而影响在实际网络中的应用。为了克服似然估计引起的问题,提出了一种新的估计方法,该方法相对于似然估计,只需要简单的数值计算,计算量较小,且计算量不会随网络规模的增加而急剧增加。通过仿真比较可以看出估测的结果接近于真实值,能够真实反映网络报文丢失的趋势,在实际网络环境中有应用价值。

Maximum likelihood estimates were often used in the network tomography to identify the loss rate. The time spent on the estimation increased sharply with the size of the network. To overcome the problems caused by MLE, a fast and simple approach was proposed to estimate loss rates. Compared with the previous methods, the proposed one only needed simple arithmetic calculation to determine loss rates, which saved more time than the MLE, and the time spent on the inference do not increased sharply with the size of the network. Through comparison and simulation, it is obtained that the loss inferences match the true results perfectly, and correctly show the loss trend. So the proposed method is very promising in the real network.