Semi-partitioned scheduling of mixed-criticality system on multiprocessor platforms

Many functions are now integrated into one computing platform to make full use of limited resources in real-time embedded systems. But most integrated functions do not have the same criticality. So many scheduling algorithms of mixed-criticality system have been proposed and studied. When all HI-criticality tasks meet deadlines in a multiprocessor mixed-criticality system, how to maximize the system resource utilization is a challenging work. Although global and partitioned mixed-criticality algorithms widely studied, they still have some shortcomings. For the purpose of finding a balance point between partitioned scheduling and global scheduling in mixed-criticality systems, the semi-partitioned scheduling algorithm proposed in recent years is a compromise method for real-time systems. In this paper, we propose a complete semi-partitioned scheduling algorithm for multiprocessor mixed-criticality systems to improve system resource utilization. The sufficient conditions for schedulability and speedup bound of the new algorithm are derived and proven. Compared with speedup bounds of other algorithms, the new algorithm is closer to the optimal algorithm for multiprocessor mixed-criticality system. We also propose two improved algorithms based on the virtual deadline of the task to balance the relationship between task criticality and task priority. Finally, a large number of experimental simulation results show that these new algorithms in mixed-criticality system can better improve system resource utilization compared to other algorithms.