Performance maximization of energy-variable self-powered (m, k)-firm real-time systems

Abstract

Many of today’s embedded devices, e.g. pacemakers or nodes in a monitoring sensor network, are expected to work energy perpetual, i.e. to be self-powered by energy harvesting from renewable sources. The major issue of such systems is the uncertainty of the available energy, influencing the application performance predictability. In many such applications, different performance levels are defined according to the patterns of job skipping. This paper proposes a performance maximization method for self-powered energy-intermittent (m, k)-firm systems via appropriate switching between the performance levels. To formally examine the impact of performance switch time on the performance-related criteria, we introduce the energy supply and energy demand functions. A sufficient schedulability test for one hyperperiod is also proposed for the preemptive fixed-priority as-soon-as-possible ($$PFP_{ASAP}$$) scheduling algorithm under variable-rate energy harvesting. We then propose a performance maximization heuristic, and compare its effectiveness to the optimal performance. The extensive simulations show that our proposed method is fast, whereas it effectively approximates the optimal solution.