Dynamic and Aggressive Scheduling Techniques for Power-Aware Real-Time Systems

  • Authors:
  • Hakan Aydi;Pedro Mejía-Alvarez;Daniel Mossé;Rami Melhem

  • Affiliations:
  • -;-;-;-

  • Venue:
  • RTSS '01 Proceedings of the 22nd IEEE Real-Time Systems Symposium
  • Year:
  • 2001

Quantified Score

Hi-index 0.01

Visualization

Abstract

In this paper, we address power-aware scheduling of periodichard real-time tasks using dynamic voltage scaling.Oursolution includes three parts: (a) a static (off-line) solution tocompute the optimal speed, assuming worst-case workload foreach arrival, (b) an on-line speed reduction mechanism to reclaimenergy by adapting to the actual workload, and (c) an online,adaptive and speculative speed adjustment mechanism toanticipate early completions of future executions by using theaverage-case workload information.All these solutions stillguarantee that all deadlines are met.Our simulation resultsshow that the reclaiming algorithm saves a striking 50% ofthe energy over the static algorithm.Further, our speculativetechniques allow for an additional approximately 20% savingsover the reclaiming algorithm.In this study, we also establishthat solving an instance of the static power-aware schedulingproblem is equivalent to solving an instance of the rewardbasedscheduling problem [1, 4] with concave reward functions.