Partitioned Scheduling With Shared Resources on Imprecise Mixed-Criticality Multiprocessor Systems

Both resource access protocols and real-time scheduling algorithms have been extensively studied in classic embedded real-time systems. However, there has been relatively little attention given to the resource access protocol and real-time scheduling algorithms in mixed-criticality systems. In this article, we pay attention to the problem of scheduling an Imprecise Mixed-Criticality taskset on a multiprocessor platform with shared resources. First, we propose an Imprecise Mixed-Criticality with MSRP (IMC-MSRP) resource access protocol, which ensures mutually exclusive access to shared resources for tasks. Secondly, we propose the schedulability test based on IMC-MSRP for a given task-to-processor mapping method. Thirdly, we propose a feasible task-to-processor mapping algorithm called RA-CU-WFD, which first assigns tasks sharing the same resources to the same processor to reduce the global waiting time of tasks and thus improve the schedulability ratio of the system. And then assigns tasks based on the Criticality-Unaware Worst-Fit Decreasing algorithm. Finally, we conduct experiments using synthetic tasksets, and the experimental results show that RA-CU-WFD outperforms other approaches in terms of the schedulability ratio.