Two basic factors influence the design of highly reliable computer systems: the amount of failures required to be tolerated and the reliability or MTBF required. A computer system designed to tolerate any three single failures in a fail-operational–fail-operational–fail-safe manner for a real-time control application is presented. The design approach uses adaptive majority voting in both hardware and software with a four-level redundant system. Various methods of performing the adaptive majority voting functions were evaluated with the selected approach using a special module termed a voter–comparator switch (VCS). The VCS module allows the computer system to be operated in a variety of redundant modes, depending on the failure tolerance required at any particular time.

The design of a computer organization is presented for general-purpose computing with a very high tolerance of failure while taking advantage of future large-scale integration techniques. A high degree of parallel computations may be executed. The organization distributes processing logic along with memory to form "cells." The architecture and operation of the cells are developed. Design aspects of the cells are presented along with software considerations of the organization.