On Design and Practice of Fault-Tolerant Real-Time Computer Systems

Abstract

One major group of real-time computer systems are those specialized in supervisory control of several highly complex systems. Computer systems used in such a supervisory control role tend to be tightly coupled with respect to time to the other complex systems and need to operate under very strict response times imposed by the problem domain. The main task for the designer of such a computer system is the proper mapping of currently available technologies to the computer system design under pre-established financial, performance and reliability constraints. Clearly, this task can only be accomplished by acknowledging the limitations, resource requirements, and characteristics of the available technology relative to the dynamics of the supervisory control time lines. Furthermore, most supervisory control systems are developed under an evolutionary approach of separate independent developments of the controlled components. The compatibility of the supervisory system with new components as well as the backward compatibility with previous versions of the supervisory system are major design issues. Typically, the life cycle for the controlled components as well as the supervisory control system itself are measured in decades.