Computers that are `never¿ down: Fault-tolerant systems based on multiple microprocessors trade off throughput for enhanced reliability

A new approach to fault-tolerant computer design-systems based on multiple microprocessors-is winning followers. This new approach is contrasted with the systems produced by Tandem Computers Inc., which are based on CPUs made up of MIS and LSI circuits. The architecture of the multiple microprocessor approach is described, with a distinction made between tightly coupled and loosely coupled architectures. Three tightly coupled systems are described to illustrate how both hardware and software can be used in fault-tolerant computers to isolate system faults and how error detection and system recovery can be carried out both by hardware and by software. It is noted that in most loosely coupled fault-tolerant systems, the emphasis is on software.