Abstract
Energy has become a first class design criterion in many of today's real-time embedded systems which are often operated by limited energy sources like batteries. Processor slowdown (using DVFS) and processor suspension/sleep (using DPM) are the two primary techniques adopted to control energy dissipation at the operating system level. This work proposes a combined slowdown-sleep strategy to minimise energy dissipation in hard real-time multicore systems, scheduled using the DPFair technique. DPFair is an important optimal proportional fair scheduler which ensures lower design costs through full resource utilisation while guaranteeing the satisfaction of all deadlines, even for dynamically arriving task sets. While there exists a DVFS based state-of-the-art energy-aware DPFair scheduler, our proposed scheduling technique, DPFair with slowdown and suspension (DPFair-SS), exhibits appreciable performance gains over the existing strategy in situations when the system experiences low workloads over significantly long durations.