An Artificial Neural Network Approach to Performability of Multiprocessor Interconnection Networks

Abstract

Performability of an interconnection system depends upon the failure characteristics of its components. characteristics of its components. There is the need of a technique to predict the per t the performability of a multipr multiprocessing network from the existing available input/output data. In an interconnection network, the processors are connected with each other through links. ough links. There may be imperfection at the links or tion at the links or at the nodes, which affect s the system performance. Hence a general and flexible prediction model needs to be developed to compute the reliability and per y and performance of the multiprocessor interconnection networks. In this paper we presents an artificial neural network model based on principle of back propagation algorithm to compute the per the performability of crossed-cube and star graph multiprocessor and star graph multiprocessor interconnection networks. interconnection networks