Cost effectiveness analysis of different fault tolerance strategies for hypercube systems

Abstract

A general model of a multicomputer system is presented, and an index is calculated for measuring its performance. Since the index depends on the average internode distance, recursive expressions for the evaluation of this distance are derived. A cost-effectiveness figure of merit is defined on the basis of a performability model that uses the performance index. The figure of merit is used to compare the cost effectiveness of two different strategies, the degradable hypercube and the nondegradable hypercube. The fault model adopted is based on the assumption of partial node failure (loss of the main processor but not of the communication coprocessor). Under this assumption, the result suggests that the degradable hypercube strategy is more cost effective.<>