Authors : Sangchul Han, Minkyu Park and Woojin Paik
Abstract: Under heavy workload mobile real-time systems make full use of system resources to meet timing constraints of tasks. Under light workload they need to reduce energy consumption due to limited system resources and a limited power source. EDZL (Earliest Deadline until Zero Laxity), one of multiprocessor real-time scheduling algorithms can enable higher system utilization than EDF (Earliest Deadline First) and its scheduling overhead can be restricted. However, there is little work on reducing energy consumption in EDZL scheduling. This study proposes a DVFS (Dynamic Voltage/Frequency Scaling) technique in EDZL scheduling. The technique includes two algorithms: one that calculates a uniform speed on full-chip DVFS platforms and the other that calculates task’s individual speed on per-core DVFS platforms. We present simulation results that show our technique is simple but effective for reducing energy consumption in real-time scheduling.
Sangchul Han, Minkyu Park and Woojin Paik, 2019. Dynamic Voltage/Frequency Scaling for EDZL Scheduling in Multicore Real-Time Systems. Journal of Engineering and Applied Sciences, 14: 8039-8046.