Improved task partition based fault-tolerant rate-monotonic scheduling algorithm

Abstract

Real-time fault-tolerant scheduling algorithms aim to reduce hardware redundancy for hard real-time systems. In this paper, improved algorithms TPFTRMS and TPFTRMG for preemptively scheduling periodic tasks with primary/backup fault-tolerant scheme are presented. The algorithms use the task partition, backup overlapping and backup deallocation techniques. Task partition makes the algorithm less sophisticated and more efficient. Backup overlapping reduces the redundancy introduced by active backup and makes the computing resources fully utilized. Backup deallocation releases resources once the computation is over and thus reduces the system overhead. Different scheduling strategies are adopted in different scenarios, which makes the algorithms more adaptive. Simulation results show that compared with previous algorithms, the new schemes need less processors and lower time complexity.