Modelling correlated transient failures in fault-tolerant systems

Abstract

Massive hardware redundancy has long been proposed as a means to achieving high reliability in critical real-time control applications. However, such an approach is only effective against independently occurring failures. Environmental disturbances, such as electromagnetic noise and radiation, often give rise to correlated transient failures in redundant systems. Mere processor redundancy is ineffective against such failures, and time redundancy must be used instead. An integrated model that takes into account both hardware and time redundancy is presented.<>