多核系统中基于动态松弛时间回收的节能实时调度算法

An Energy-efficient Multi-core Real-time Scheduling Algorithm Based on Dynamic Slack Reclamation

许多嵌入式实时任务的实际执行时间往往小于最坏情况执行时间,因此可以产生大量的动态松弛时间。针对时限等于周期的偶发任务集,提出一种基于动态松弛时间回收的多核系统节能实时调度算法DSREM。该算法基于最优在线调度算法LRE-TL,利用TL面内节能实时调度思想,在每个TL面的提前完成时刻实现动态松弛时间回收,降低后续偶发任务的执行频率,实现更多的节能。同时该算法还在每个TL面的初始时刻、偶发任务的释放时刻实现动态电压和频率调节,保证了偶发任务集最优可调度性。经过系统的数学分析和仿真,结果表明,DSREM算法不仅保证了偶发任务集的最优可调度性,而且当任务集总负载超过某一个值后,其节能效果始终优于现有方法,尤其随着总负载的增加,节能效果会更好。

In many embedded real-time systems, the actual execution time of tasks is usually less than their worst-case execution time （WCET）, which produces lots of dynamic slack time. Based on this fact, we proposed an energy-efficient real-time scheduling algorithm LRE-DVFS-ALL for sporadic tasks deployed on multi-cores, which is based on optimal online scheduling algorithm LRE-TL. The main idea of the algorithm is to reclaim dynamic slack time, during which DVFS techniques can be used to reduce the execution frequency of future tasks to eliminate the energy consumption. Meanwhile, our algorithm also dynamically scales the voltage and frequency at the initial time of each TL plane and the release time of a sporadic task in each TL plane to guarantee the optimal schedulability of sporadic tasks. Systematic mathematical analysis and extensive simulation results show that LRE-DVFS-ALL can always save more energy than the existing algorithms when the total workload of the system exceeds a threshold, and it can also guarantees the optimal sehedulability at the same time.