An Improved DVFS Algorithm for Energy-Efficient Real-time Task Scheduling

摘要

Power consumption is one of the most critical challenges in real-time systems. Dynamic voltage and frequency scaling (DVFS) has been widely used to balance performance and energy consumption. We first analyze two well known DVFS scheduling algorithms: CC-EDF (Cycle-conserving earliest deadline first) and LA-EDF (Look-ahead earliest deadline first). Then, we find out that LA-EDF may provide illegal frequency (higher than system capability). By addressing this issue, we develop LA2-EDF which only considers future task load and always sets a high frequency at end. Based on CC-EDF and LA2-EDF, we propose a novel algorithm named Soft-LA2 by considering both current and future task load. Simulation results show that under certain parameter settings, Soft-LA2 can produce energy-saving up to 28% than LA2-EDF. Furthermore, we analyze the relationship between CC-EDF, LA2-EDF, and Soft-LA2 and provide advice for Soft-LA2’s trade-off parameter settings.