NetTomo: A tomographic approach towards network diagnosis

Network diagnosis is a vital aspect in ensuring an efficient and robust functioning of any kind of mesh network. In this paper we present a network diagnosis method which determines the delay map of a mesh network using only end-to-end delay measurements without having the knowledge of the path taken. We model the problem of network diagnosis as an inverse problem and using a concept of ray tracing, solve for the delay in the network. With the help of simulations we show that our algorithm is able to detect nodes in the network based on their delays with reasonable accuracy using only O(n) probes for obtaining measurements. We further demonstrate a real world application of our algorithm in the domain of internet backbone networks by using data pertaining to a major US based network provider.