Directed acyclic graph scheduling for mixed-criticality systems

Abstract

Deploying safety-critical systems into constrained embedded platforms is a challenge for developers who must arbitrate between two conflicting objectives: software has to be safe and resources need to be used efficiently. Mixed-criticality (MC) has been proposed to meet a trade-off between these two aspects. Nonetheless, most task models considered in the literature of MC scheduling, do not take into account precedence constraints among tasks. In this paper, we propose a multicore scheduling approach for a model presenting MC tasks and their dependencies as a Directed Acyclic Graph (DAG). We also introduce an evaluation framework for this model, released as an open source software. Evaluation of our scheduling algorithm provides evidence of the difficulty to find correct scheduling for DAGs of MC tasks. Besides, experimentation results provided in this paper show that our scheduling algorithm outperforms existing algorithms for scheduling DAGs of MC tasks.