Feedback EDF Scheduling Exploiting Dynamic Voltage Scaling

Dynamic voltage scaling (DVS) is a promising methodfor embedded systems to exploit multiple voltage and frequencylevels and to prolong battery life. However, pureDVS techniques do not perform well for systems with dynamicworkloads where the job execution times vary significantly.In this paper, we present a novel approach combiningfeedback control with DVS schemes targeting hardreal-time systems with dynamic workloads. Our method reliesstrictly on operating system support by integrating aDVS scheduler and a feedback controller within the EDFscheduling algorithm. Each task is divided into two portions.Within the first portion, the objective is to exploitfrequency scaling for the average execution time. We reserveenough time for the second portion to meet the deadlinerequirements up to the worst-case execution time followinga last-chance approach. Feedback techniques makethe system capable to select the right frequency and voltagesettings for the first potion, as well as guaranteeing hardreal-time requirements for the overall task. Simulation experimentsdemonstrate the ability of our algorithm to saveup to 29% more energy than previous work for task sets withdifferent dynamic workload characteristics.