Reliability comparison of computer networks

Abstract
The authors consider networks modeled as directed graphs with a uniform link failure probability (q) and develop an algorithm which permits not only a symbolic reliability evaluation, but also a comparison of different topologies to verify if a particular one is uniformly maximally reliable, i.e. it has the best reliability for any value of q. In the case when the previous property cannot be verified, it is possible to evaluate exactly the range of values of q for which one network is better than the other. The method can be extended also to compare networks with different topological characteristics, namely different number of links and/or nodes.