Fault Tolerant DMR Miroprocessor System

Abstract

Conceptually, duplication with comparison is quite efficient fault detection technique. The simplicity, low cost, and low performance impact of the comparison technique are particulary attractive. In the paper we discuss the problem of designing double modular redundancy systems (DMR’s) tolerating permanent and transient faults. The last class of faults is quite critical, however it was neglected in the literature. Analyzing DMR system performance we introduced some parameters which specify transient fault tolerance capabilities. This analysis allowed to outline effective methods of dealing with transient faults.