Chinese Journal of Computers   Full Text
Title	A Hybrid Static/Dynamic Energy-Aware Weakly-Hard Real-Time Scheduling Algorithm Based on Simple Feedback
Authors	WU Tong1)，2) ZHANG Dong-Song1) JIN Shi-Yao1)
Address	1)(National Laboratory for Parallel and Distributed Processing, Institute of Computer, National University of Defense Technology, Changsha 410073) 2)(School of Continuing Education, National University of Defense Technology, Changsha 410073)
Year	2009
Issue	No.6(1140—1146)
Abstract &   Background	Abstract With the emergence of the prominent problem of energy consumption, energy efficient real time scheduling is hot. Based on the Worst Condition Execution Time (WCET), the hybrid static/dynamic algorithm calculates the processor speeds offline, which results in the limited energy efficiency. To solve this problem, a new algorithm is proposed, in which, the simple feedback mechanism is introduced to estimate the actual execution time. The whole speed is decreased by splitting the job into two parts. The execution time is prolonged, and so the new algorithm is more energy efficient. The experimental results show that the proposed algorithm outperforms the original one when the Average Case Execution Time (ACET) is much less than the WCET, which can improve energy savings about 60% to 70% at most and about 10% at least. Unfortunately, when the ACET is close to the WCET, the proposed algorithm consumes more energy than the original one. Keywords real-time scheduling; energy efficient; feedback; weakly hard real-time Background Nowadays, energy-aware scheduling is attracted in the research of real-time systems, in which Dynamic Voltage Scaling (DVS) technique is generated from hardware based on low-power design and has been one of key technologies in real-time system. DVS can be easily included into real-time scheduling methods to make energy savings by scaling the voltage and frequency while maintaining real-time deadline guarantees. Currently, most existing energy-aware scheduling algorithms are focused on the hard real-time system. However, many practical real-time applications exhibit more complicated characteristics, which is called the Quality of Service (QoS) requirements. For example some applications may have soft deadlines where tasks which do not finish by their deadlines can still be completed with a reduced value; or they can simply be dropped without compromising the desired QoS levels. The techniques based on the traditional hard real-time systems become inefficient of inadequate when QoS requirements are imposed on the systems. And so the energy-aware weakly-hard real-time scheduling algorithms are introduced to solve this problem. In this paper, after analyzing the shortages of the hybrid static / dynamic algorithm, we proposed a new algorithm, in which, the simple feedback mechanism is introduced to estimate the actual execution time. The whole speed is decreased by splitting the job into two parts. The execution time is prolonged, and so the new algorithm is more energy efficient. Authors have done research on weakly-hard real-time system for many years. By these studies, they have a deep understanding of weakly-hard real-time system and publish many papers in this area. This paper contributes on the field of energy-aware weakly-hard real-time scheduling to get more energy savings. The experimental results show that the proposed algorithm outperforms the original one when the Average Case Execution Time (ACET) is much less than the WCET, which can improve energy savings about 60% to 70% at most and about 10% at least. Unfortunately, when the ACET is close to the WCET, the proposed algorithm consumes more energy than the original one.