The development of fault tolerant computer systems using dual processing techniques

Abstract

The paper discusses ways of building a dual processed computer system to provide a highly reliable fault tolerant configuration. The need and justification for fault tolerant systems is discussed and the comparative availability that can be expected from single and dual configurations is calculated. Based on systems implemented by CAP, two of the most important types of dual processing systems are discussed: namely load sharing and continuous processing systems. The load sharing approach to fault tolerance is particularly important for transaction processing applications because it offers the opportunity to build a fault tolerant system for a cost close to that of a conventional single-computer system. An example based on CAP's system for the Durham Constabulary on Perkin-Elmer equipment is discussed. The continuous processing system described is based on CAPTEC, CAFs operating system for Digital Equipment Corporation computers, which provides a continuous processing system where the system continues uninterrupted on a single failure.