Resource-Aware Task Allocation on Mixed-Criticality Systems: a Task-Splitting Approach
Ruoxian Su, Hanzhi Xu, Jieyu Jiang, Shuai Zhao · 2024
An important trend of real-time systems is to integrate applications with different criticality levels on a single multicore platform, enabling resource sharing among applications . However, the existing task allocation schemes suffer from the issue of severe resource contention between cores for accessing mutually exclusively shared resources, resulting in significant blocking time. This jeopardizes the system schedulability and leads to the application difficulty of mixed-criticality systems (MCS) in real-world systems. To tackle this issue, this paper proposes a resource-aware task allocation (RATA) algorithm for multicore mixed-criticality systems. The proposed allocation takes the resource usage of tasks into account and aims to localize the most frequently accessed resources by allocating the requesting tasks on the same core, which effectively reduces the inter-core resource contention, hence, improving system schedulability. In addition, a specialized allocation process is constructed for different execution modes in MCS with a task-splitting mechanism, enabling direct support of MCS with improved resource utilization. The experimental results show that RATA outperforms existing methods by 71.18% on average (up to 217.85%) in terms of system schedulability.