Fault Tolerant Global Scheduling for Multiprocessor Hard Real Time Systems

In hard real-time systems, faults happen every now and then. With respect to the criticality of this kind of system, the capability of fault tolerance is necessary. A fault tolerant scheduling algorithm is capable of guaranteeing timing requirements of tasks even when faults occur. A primary-backup based fault-tolerant global scheduling algorithm RRFTGS (resource reclaim fault tolerant global scheduling) is proposed in this paper. RRFTGS pushes the execution of backup as late as possible and reclaims the resource distributed to backups when it is no longer needed. By this method the additional cost for achieving fault tolerance is significantly reduced. Simulation result shows that, comparing to the scheduling algorithm which only use passive backups, RRFTGS needs much less processors to tolerate a single fault. RRFTGS also could be adopted when high utilization task exists.