Highly reliable computers are needed to support execution of safety critical software. Modern VLSI technology, coupled with advanced Fault tolerance techniques, offers the potential of a computer with a predicted reliability of .9999 over a 5 year operating cycle without repair. Using voting techniques to detect and mask the effects of both...and transient failures, it appears practical to reduce the probability of an unprotected single bit error to less than lx10-9 over a 5 bit year cycle. The proposed effort will research the feasibility of designing such a system using currently available VLSI modules and field-proven fault tolerance approaches. This research effort will include: (1) exploration of the requirements of safety critical computer applications in both government and industry; (2) analysis of the trade offs and relationships among component reliability, integration, cost, and fault tolerance techniques. A architectural definition of a fault tolerant computer using currently available component and fault tolerance technology; and (4) evaluation of the design for feasibility in terms of reliability, performance, and cost. August systems brings to this effort practical experience with commercial fault tolerance computers in a wide range of safety critical control applications.
