Partitioned EDF scheduling in multicore systems with quality of service constraints

In this paper we study the partitioned EDF scheduling in a homogeneous multiprocessor environment with Quality of Service (QoS) constraints. The system considered here is a real-time multiprocessor system assumed to be powered by rechargeable batteries. We address the issue of how to best partition a set of firm real-time tasks that can occasionally skip one instance according to a predefined QoS threshold. The main goal is to minimize the energy consumption of the system while offering solutions with respect to transient energy starvation situations the system can experiment. The contribution of the paper is twofold. First, we present a schedulability analysis of firm multiprocessor task sets under QoS constraints. Second we propose new partitionning heuristics integrating skips. The evaluation is conducted from several points of view (minimization of the total processor number, maximization of the spare capacity on each processor).