No abstract

Scheduling of mixed-criticality systems (MCS) on a common computational platform is challenging because conventional scheduling approaches may cause inefficient utilization of shared computing resources. In this paper, we propose an approach called Clustering-based Partitioned Earliest Deadline First (C-PEDF) algorithm to schedule dual-criticality implicit-deadline sporadic tasks on a homogeneous multicore system. Our C-PEDF scheduling approach exploits (i) a Clustering-based bin-packing algorithm that explicitly accounts the demands of tasks based on their levels of confidence; and (ii) an Enhanced dual-mode scheduling policy to schedule tasks within a core. The proposed C-PEDF integrates every single high-level workload with a group of low-level workloads and coalesces them into a cluster. Within each cluster, tasks are scheduled under our Enhanced dualmode scheduling policy to improve the service level of high-level tasks without jeopardizing the schedulability of low-level tasks. Clusters are scheduled under Earliest Deadline First (EDF) scheduling approach. We conduct a schedulability test for the proposed technique, and we demonstrate how workloads can be clustered by means of Mixed Integer Nonlinear Programming (MINLP) model. Extensive simulation results reveal that our algorithm significantly outperforms other existing approaches both in acceptance ratio and the impact factor of low-level tasks.

No abstract