Reliability Research Of Some Interconnection Networks

The processors in multiprocessor computer systems are connected through various interconnection networks.Typically,the topology of an interconnection network is often modeled by a connected graph G where the vertices correspond to processors and the edges correspond to the communication links between the processors.One fundamental criterion in the design of interconnection networks is reliability.There are many methods for measuring reliability.In general,they can be divided into the deterministic method and the probabilistic method.This article researches the reliability of interconnection networks from these two methodsFirst,the reliability of the interconnection network is described from a deterministic method,that is,the reliability of the system is measured by the(edge)connectivity of the interconnection network topology.For further research,the concept of conditional(edge)connectivity is proposed,that is,after some vertices(edges)are deleted,the remaining subgraph still retains some property,such as component(edge)connectivity,extra(edge)connectivity.In second chapter,we determine the(s+2)-component connectivity of the generalized exchanged hypercube GEH(s,t),the extra edge connectivity and the component edge connectivity of the augmented cube AQn.Due to technological advances,multiprocessor systems may consist of hundreds or thousands of processors.With the growth of the system,some of the processors may fail while the system is running.Quantifying the impact of the failure is important.So we can use the probability method to study the reliability of the system,such as the fault tolerance of the system can be characterized by the probability that a subsystem of a certain size can remain connected after the failure of the interconnection network,that is,the subsystem(subgraph)reliability.In third chapter,we calculate the estimated values of subgraph reliability of the dual cube Dn and the balanced hypercube BHn under the probability fault model.