Assessing the reliability and safety of fault tolerant designs

Present-day fault-tolerant systems are increasingly used in critical applications such as aircraft control systems and industrial controllers. Dependability considerations require that in addition to high reliability, the system must have a high level of safety. Briefly, the reliability R(t) is the probability that the system produces correct output. Safety S(t) is defined as the probability that the system output is either correct or the error in the output is detectable. Fault tolerance is implemented through hardware redundancy. In this architecture, a functional unit is replicated N times, forming an N modular redundant system. Each unit produces a single bit output and the results are applied to a majority voter. This paper presents a framework for considering both reliability and safety of NMR systems, estimations of R(t) and S(t) functions and trade-off considerations