Energy Optimized Non-preemptive Scheduling of Real-Time Tasks with Precedence and Reliability Constraints

Abstract

The contemporary applications have become highly parallel and computation-intensive, which can be processed efficiently on a distributed heterogeneous multiprocessor system. However, the consequences such as precedence constraint must be taken care of by the scheduling mechanism. In addition, the lower energy consumption and higher reliability are among the crucial, however conflicting system design objectives for the real-time systems. In this work, we address the problem of scheduling of a set of non-preemptive real-time tasks with precedence, timing, and reliability constraints on a heterogeneous multiprocessor system to optimize energy consumption. In addition, we consider the effects of waiting time on the system reliability, which is inescapable, especially for parallel applications. Motivated by the intractability of computing an optimal solution, we propose a simulated annealing-based approach to obtain a near-optimal solution to the aforementioned problem. Additionally, a heuristic is also presented to compute a solution in a reasonable time. We perform an extensive simulation study for the synthetic as well as practical application graphs and report the performance with respect to various simulation parameters.