A MULTIPROCESSOR BASED FAULT TOLERANT REAL TIME TASK SCHEDULING ALGORITHM

Fault tolerance is the important requirement of real time systems. In real time systems, if there is a real time task that does not finish before its deadline, then system will appear abnormal. In this paper, a fault tolerant scheduling algorithm applied for multiprocessor real time system is proposed. Primary backup technique and first fit heuristic method are adopted in this algorithm. The algorithm provides fault tolerance through reserving the re execute time for the failed tasks caused by processor failure, and improves the processor utilization and the task accepted ratio through reclaiming and re assigning the reserved time to other tasks if there is no task failure, as well as through overlapping the reserved time. Simulation results show that the algorithm is efficient.