REAL-TIME PROGRAMMING WITH FAULT-TOLERANCE

Real-time programming is one of the most difficult tasks in software engineering. With advancing technology in high speed computer hardware and increasing complexity of software systems, the current trend is to develop high level programming languages to support real-time systems. Moreover, a fault-tolerance approach is required in many real-time systems to achieve high reliability.; The major goal of the research is to develop methodologies and mechanisms for high level systems programming languages which facilitate programming real-time systems and fault-tolerant real-time systems.; Some language requirements for fault-tolerant real-time programming are identified. Four existing languages, HAL/S, Modula, concurrent Pascal, and Path Pascal, are evaluated against each requirement. Path Pascal, a locally developed system programming language, has been chosen as a vehicle to experiment on proposed methodologies and mechanisms for programming fault-tolerant real-time systems.; Recovery blocks and a deadline mechanism have been incorporated into the Path Pascal language to aid fault-tolerant software design in real-time systems. Facilities for specification of timing constraints and estimate of execution times are provided in the language. The verification of meeting deadlines subject to the timing constraints for some selected scheduling algorithms is also available during compilation.; A fault-tolerant real-time executive and a simulation of a satellite on-board computer system have been programmed using deadlines in Path Pascal. The results show the potential of the fault-tolerant mechanisms in a high level language to describe time-critical real-time systems. The domino-effect problem has been discussed for a periodic real-time system. A subsystem in the simulation has been shown to be domino-free.; Finally, a reliability model for a fault-tolerant, periodic real-time software system is presented. The system is constructed using the fault-tolerant approach complemented with fault-avoidance techniques. The reliability model supports the application of fault-avoidance techniques to the construction of fault-tolerant real-time systems. The model also suggests that a majority voting scheme be used to complement the recovery block approach in order to achieve a desired reliability. The results of this research definitely extend the area of fault-tolerant real-time system programming.