A new fault-tolerant scheduling technique for real-time multiprocessor systems

Abstract

In multiprocessors used for implementing real-time applications, it is strongly required to ensure that tasks' deadlines are met in spite of faults. In this paper, we discuss fault-tolerant task scheduling techniques in real-time multiprocessor systems where aperiodic tasks arrive dynamically. For fault-tolerance, multiple copies of each task have to be scheduled on distinct processors in such systems. The authors propose a new technique for efficiently scheduling copies of tasks having small laxity. Then, we integrate our technique with previous work and develop a new scheduling algorithm, removing the assumption on task laxity. Simulation results show the effectiveness of our technique.