A case study on the fault-tolerizability of an archetypical shared memory minisuper multiprocessor

Issues in assessing fault-tolerance are addressed. A specific archetypical high-performance shared-memory architecture is chosen as a test case (specifically the Encore Multimax, to be used in space defense applications and environments), and a set of target fault-tolerance requirements are specified. The requirements are established to meet realistically the anticipated needs of aerospace applications of the 1990s. A methodology is then presented to systematize the steps and document the alternatives to be evaluated and the issues to be resolved in the process of fault-tolerization. The design of the resultant final target fault-tolerant machine is presented. The advantage to design based on parallel architectures is that resources can be dynamically reallocated to enhance either performance or fault-tolerance.< >