Affiliation:
1. Huazhong University of Science and Technology, Wuhan, China
Abstract
In order to save the energy consumption of real-time embedded systems, the integration of Dynamic Voltage and Frequency Scaling (DVFS) and Device Power Management (DPM) techniques has been well studied. In this article, we propose a new energy management scheme for periodic real-time tasks with implicit deadlines. We mainly focus on the DPM part by presenting a novel approach to the real-time DPM problem. Specifically, we first identify intervals for each device, which we refer to as Crenel Intervals, by partitioning the Earliest Deadline First (EDF) schedule of the tasks that need to access the device into successive intervals. The principle for identifying Crenel Intervals is that for each task, there is only one deadline located in each Crenel Interval. Next, targeting at a single device model and a multiple device model, respectively, we propose the CI-EDF and CI-EDF
m
algorithms to schedule task instances in each Crenel Interval, so as to form long and continuous slacks in each Crenel Interval but without jeopardizing any task deadlines. Then, the slack in the Crenel Intervals can be utilized to perform not only DPM, but also DVFS. The experimental results show that our approaches can achieve considerably more energy savings than existing techniques with comparable quality.
Funder
National Natural Science Foundation of China
China Postdoctoral Science Foundation
State Key Program of National Natural Science of China
Publisher
Association for Computing Machinery (ACM)
Subject
Hardware and Architecture,Software