Method for estimating delays in parallel redundant data transfer networks

Abstract

This paper presents a unified method for estimating latencies in a parallel redundant network with losses. The proposed method is based on mathematical statistics and, in the opinion of the authors, is more instructive in small networks than typical estimations that are based on the Markov chain model, which is recommended by ITU. The method requires known CDFs of latencies of parallel networks, as well as their packet loss rates. The CDFs can be built from experimental data or can be approximated in accordance to assumed distribution function. The method eliminates the need to calculate the transition probabilities that are required in the Markov chain. Also, the method can be used for various number of parallel networks. However, the method is applicable only for independent networks