Evaluation of fault tolerant software structures for parallel systems in industrial control

Computing systems are being used increasingly in applications which demand safe and reliable operation at all times. However, fault-free operation under all operational circumstances cannot be guaranteed. Fault tolerance is achieved by including in the system design special mechanisms which recognise and override automatically the effects of the fault, giving a satisfactory performance even in the presence of the fault. In the paper these software fault tolerant structures are applied to a real-time robotics application distributed over a number of transputers. The design and implementation language is Occam. The effectiveness of these structures against faults is measured both in terms of data integrity and performance aspects. The paper shows the effectiveness of certain software fault tolerance mechanisms when applied to a parallel system, with real time responses.<
>