| Popis: |
Virtualization plays a crucial role in successfully integrating multiple real-time applications on a single platform. Through virtualization, different applications can run in independent virtual machines, isolated from each other to prevent mutual interference. This paper considers partitioned fixed priority (P-FP) scheduling for tasks and servers on the multi-core virtualization platform with non-preemptive resource sharing. The solution for P-FP scheduling on multi-core virtualization environments typically uses vMPCP, an extension protocol of the well-known Multiprocessor Priority Ceiling Protocol (MPCP), to address the problem of resource sharing among real-time application tasks. However, the existing analysis of Worst-Case Blocking Time (WCBT) for the P-FP+vMPCP scheduling tends to be overly conservative, resulting in many schedulable task sets being incorrectly classified as unschedulable. Consequently, it leads to wasted system resources and degraded performance. To mitigate this issue and reduce the pessimism of existing analysis, we propose an improved WCBT analysis. This analysis bounds the maximum blocking time that tasks experience when each job from the same task requires to access the same type of non-preemptive global resource multiple times. As a result, it enhances the accuracy over prior work and obtains more precise Worst-Case Response Time (WCRT) for tasks. In our experiments, the WCRT-based schedulability test is adopted to evaluate the schedulability of tasks. Additionally, we do not consider additional scheduling costs, as no online decisions are required. Experimental results demonstrate that our proposed method outperforms the latest advanced methods in terms of the schedulability rate of tasks. Specifically, the schedulable task sets are increased by approximately 47.7% on average. |
