End-to-end path loss inference algorithm with network tomography

Network path loss rate is an important indicator of the network performance. Given a network with n end hosts, existing systems require O(n log n) measurements, and thus consume more management time and costs. An efficient network path loss rates inference method is proposed in this paper, which only needs to measure less part of paths to infer the loss rate of all remaining paths with higher accuracy, and ensure the quality of monitoring and reduce the management cost. The simulation results show that our method saves 5% ∼ 14% probing paths than the existing method. Moreover our algorithm reduces the management cost of the network monitoring to some extent.