Files
Abstract
Computer systems are prone to faults. Faults are of two types: transient and permanent. The performance of events becomes unpredictable upon the occurrence of such faults. This behavior introduced by transient faults might cause deadlines to be missed making hard real-time systems fail. Checkpointing is a cost-effective way of detecting faults. A generalized multi-frame (GMF) task is a sporadic task model where the execution times, deadlines and minimum separation times are all N-array vectors. Upon an occurrence of a fault, the set of GMF tasks re-execute causing an increase in the overall execution time of the job causing other jobs to miss their deadlines. To make the system schedulable, this thesis introduces a checkpointing mechanism called Generalized Multi-frame with Fault Tolerance (GMFFT) for the GMF tasks. We tackle faults by checkpointing the set of GMF tasks and find the worst-case recovery time sequence of the given task set.