A design approach for ultrareliable real-time systems

A design approach developed over the past few years to formalize redundancy management and validation is described. Redundant elements are partitioned into individual fault-containment regions (FCRs). An FCR is a collection of components that operates correctly regardless of any arbitrary logical or electrical fault outside the region. Conversely, a fault in an FCR cannot cause hardware outside the region to fail. The outputs of all channels are required to agree bit-for-bit under no-fault conditions (exact bitwise consensus). Synchronization, input agreement, and input validity conditions are discussed. The Advanced Information Processing System (AIPS), which is a fault-tolerant distributed architecture based on this approach, is described. A brief overview of recent applications of these systems and current research is presented.<<ETX>>