Energy-Aware Scheduling of Execution-Instant Sensitive Real-Time Systems

Energy-Aware Scheduling of Execution-Instant Sensitive Real-Time Systems

Abstract

Time constraints in real-time systems are traditionally identified based on the job completion-times, i.e. using classic deadlines or time/utility functions (TUFs). In this paper, we consider a new model of time constraints, namely Instant-Value-Function (IVF) which specifies the most appropriate instants of time to execute a job. More precisely, we see that in many applications in the embedded world, the exact instants of time within which a job is executed affect the value that the job accrues. This property cannot be expressed by classic deadlines or TUFs. In this paper, we consider simultaneous timeliness and energy optimization of battery-operated IVF-constrained real-time embedded systems. For energy optimization, we discuss the semantics of applying dynamic voltage scaling (DVS) to such energy-bounded systems. We see that, in spite of the completion-time dependent time constraint models, the shape of job IVF is affected by changing the speed of the system processor. Accordingly, we present a DVS-based scheduling algorithm to maximize the accrued value of the energy-bounded real-time systems. The optimality of this algorithm is established analytically, while its effectiveness is also confirmed through simulation experiments.

References