复杂实时系统常采用资源分区隔离软件子系统,分区结构的系统包含分区级和任务级2层调度。抢占需要消耗系统资源,过多地抢占会降低系统效率。针对该问题,文章提出了延迟抢占分区模型,在分区中设置不可抢占区,保证系统可调度性的同时减少了大量不必要抢占;建立了延迟抢占分区模型的可调度性判定条件,并给出了分区设计算法。仿真结果表明,延迟抢占分区可以显著减少运行中的抢占次数。
For security concern, resource partitioning is adopted in complex real-time systems. The partitions are scheduled by a system-level scheduler while the tasks are scheduled by a task-level scheduler which resides in each partition. The preemptions which are brought about by the schedulers cause considerable run-time overhead, and a fairly large number of them are not necessary with respect to the schedulability. In this paper, a deferred-preemption partition model is proposed to address this problem. The model sets up non-preemptive regions in each partition to reduce unnecessary partitionlevel preemptions, as well as the preemptive overhead. And the schedulability test based on the resource supply and demand function is established. The partition design is also presented. The experimental result shows that the proposed model can reduce significant preemptions.