Checkpointing for single core energy-neutral real-time systems

In this work, we focus on energy-neutral real-time systems, where ambient energy harvested from the environment powers a device that executes tasks with timing constraints.

We address the scheduling of a set of real-time sporadic tasks with predefined checkpoints on a single-core platform powered by an energy harvester, using the Earliest Deadline First (EDF) scheduler. Our objective is to minimize temporal and energy checkpointing costs by selecting a subset of checkpoints per task while ensuring compliance with both timing and energy constraints.

We propose a novel energy-aware scheduler and two techniques for selecting checkpoints that respect both energy and timing constraints. We provide a detailed schedulability analysis for the proposed techniques and present results from a large set of synthetic experiments. These experiments demonstrate that our approaches outperform ED-H , an optimal energy-neutral preemptive scheduler in average by 50%.