Online DVFS Scheduling Algorithm for Sporadic Tasks with Energy-Conscious in Hard Real-Time Systems
Manal A. El Sayed, M. Saad El Sayed, Shahira M. Habashy, Rasha F. Aly · 2021
Dynamic voltage frequency scaling (DVFS) has been an efficient technology in minimizing the energy consumption of real-time embedded systems. Many prior studies have attempted to balance performance and energy consumption through energy-efficient task scheduling using the DVFS approach. This paper presents a novel DVFS algorithm, named Energy Conscious-DVFS (EC-DVFS) that can be used with sporadic constrained-deadline hard real-time tasks implemented upon multicore platforms, where the scheduler is the Earliest Deadline First (EDF) algorithm. The EC-DVFS technique is an adaptive online speed determination mechanism that determines the safe speed for each released task firstly. That speed ensures all deadlines are achieved when the jobs are scheduled using EDF. Then, when the task is assigned to a core, the EC-DVFS adjusts the core speed online according to its actual workload by considering both current and future task load using runtime queues. We demonstrate the efficacy of the suggested approach through an extensive set of experiments. The experimental simulation results demonstrate that the suggested algorithm saves a significant amount of energy when the system is running across a variety of applications.