Overall energy reduction for hard real-time systems based on threshold work-demand analysis

Abstract

In this paper, we study the problem of reducing the overall energy consumption for hard real-time systems scheduled by the earliest deadline first (EDF) scheme. To balance the static and dynamic energy consumption, the concept of critical speed was proposed. Moreover, when combined with the processor shutdown strategy, the critical speed was widely used as the lower bound for voltage scaling in literature. In this paper, we show that the critical speed strategy might not always be more energy efficient than the traditional DVS strategy and there is a tradeoff between these two strategies depending on the job workload to be finished within certain interval. To effectively address this issue, we propose a technique that combines these two strategies to achieve better energy saving performance. Our approach determines the energy efficient speed for real-time tasks based on the threshold work-demand in a feasible interval and will reduce the job speed below the critical speed when necessary. Our experimental results demonstrate that the proposed techniques significantly outperformed the previous research in overall energy saving performance.