Temporal dependence network loss tomography using maximum pseudo likelihood method

Understanding network link loss is particularly important for optimizing delay-sensitive applications. This paper addresses the problem of estimating temporal dependence characteristic of link loss by using network tomography. First, the k-th order Markov Chain (k > 1) is introduced to model the packet loss process. The model considers the dependence of k + 1 consecutive packets, and is capable of capturing the temporal dependence characteristic of link loss accurately if k is large enough. Second, we propose a maximum pseudo likelihood inference based method to estimate the state transition probabilities of the k-th order Markov Chain link loss model from the unicast end-to-end measurements. The analytical and simulation results show the good performance of our method.