基于流量矩阵估计的路由推断算法

A Routing Inference Algorithm Based on Traffic Matrix Estimation

路由信息对网络而言非常重要。现有的路由推断技术很难适用于任意网络。本文提出了一种路由推断算法,在已知网络的拓扑结构和链路流量的情况下,先通过期望最大化算法对一个包含了多种路由可能性的流量矩阵进行估计,然后根据对应的可能需求的估计值的差异来推断实际路由。仿真结果证明了方法的有效性。

Routing information is very important to networks. The existing routing inference techniques can hardly be applied to an arbitrary network. This paper proposes a routing inference algorithm. Given the network topology and the link traffic, it first utilizes the expectation maximization （EM） algorithm to compute the traffic matrix that includes multiple routing possibilities. Then it conjectures the actual routing according to the discrepancy among the estimations of possible demands. Simulation results prove this algorithm effective.