A fault-tolerant technique for scheduling periodic tasks in real-time systems

Abstract

We present a heuristic for producing a fault-tolerant schedule of given periodic tasks in distributed real-time systems. Tasks are divided into two classes according to their task utilization. In order to recover from faults, a hybrid scheme based on space redundancy and time redundancy is used. We use a very simple and fast heuristic to provide fault tolerance and reduce time overhead in case of transient faults in distributed real-time systems. We show that our approach can improve processor utilization.