一种多点多播丢包率推断方法

A Method of Multi-point Multicast Packet Loss Rate Inference

在多点多播测量中，使用极大似然估计算法推断丢包率精确度高但复杂度也高．而使用快速算法推断丢包率速度快但准确度下降。为提高测量结果的速度和精确度，将加权平均算法MVWA（Minimum Variance Weighted Average）和快速算法u-MLE（u-Pseudo MaximumLikelihood Estimate）相结合，即根据发送源的数量进行子树划分，先利用u—PMLE算法估计各个子树的丢包率，再根据MVWA算法求出多点拓扑各个链路的丢包率。仿真结果表明，该方法适合大型网络丢包率的快速准确推测。

In multi-point multicast measurement,the maximum likelihood estimation algorithm for loss rate inference is accuracy, but with high complexity, using fast algorithm inferred with high speed but accuracy decreased. To improve speed and accuracy,the arithmetic is based on algorithms MVWA and u-PMLE. First, using u-PMLE algorithm to estimate the loss rate of multiple trees, then, the MVWA algorithm treats the trees separately and then averages the results. Simulation result shows that the method suitable for large network loss rate inference.