A Hybrid Recon guration Scheme for Real - TimeFault - Tolerant

Non-stop operation, despite the occurrence of faults, is a key requirement for highly-reliable, real-time systems , such as aircraft and nuclear reactor controller computers. Reconnguration is commonly used to mean the process of transforming the execution structure of a system from one state to another when a fault occurs. We identiied two types of reconngura-tion: fault-tolerating reconnguration (upon occurrence of each fault) and performability-enhancing reconngu-ration (independently of fault occurrences). This paper proposes a hybrid reconnguration scheme for real-time fault-tolerant systems. The proposed scheme not only meets the timing constraints or deadlines of hard real-time jobs, but also maintains high performability. In this scheme, a system reconngures itself either when a fault occurs or when a job arrives or completes its execution.