High-reliability architectures for networks under stress

Abstract

In this thesis, we develop a methodology for architecting high-reliability communication networks. Previous results in the network reliability field are mostly theoretical in nature with little immediate applicability to the design of real networks. We bring together these contributions and develop new results and insights which are of value in designing networks that meet prescribed levels of reliability. Furthermore, most existing results assume that component failures are statistically independent in nature. We take initial steps in developing a methodology for the design of networks with statistically dependent link failures. We also study the architectures of networks under extreme stress.