基于任务同步及节能的单机系统实时动态调度算法

A dynamic real-time task scheduling algorithm with task synchronization and energy saving for uniprocessor systems

提出了一种基于任务同步及节能的单机系统实时混合动态调度算法(HDSA),以有效解决能耗及实时任务同步时的优先权反转所导致的缺乏实时可调度性的问题。HDSA 包含静态算法及动态算法两部分。静态算法可在静态条件下,固定临界区的运行速度,并求出非临界区部分的静态速度。动态调度算法在实际运行中,在满足周期性任务实时可调度性及任务同步的条件下,充分利用及回收任务运行时剩余的执行时间,调节处理器的速度,以有效降低能耗。同时也能避免高优先权任务被阻塞时,低优先权任务的临界区继承高优先权任务的速度所造成的处理器电压开关的频繁切换,故能有效地降低实时任务调度的成本。实验测试表明 HDSA 在调度性能上明显优于相关的有效算法。

This paper addresses the problems of energy saving and schedulability of periodic real-time tasks in uniprocessor systems, which are caused by priority reversion as the synchronization of tasks occur. The available algorithm proposed, namely the hybrid dynamic scheduling algorithm （HDSA）, is comprised of two parts, a static and a dynamic. The static algorithm keeps the processor frequency fixed when running critical sections offline, and obtains the static speed of noncritical-section parts. The dynamic algorithm elaborates on the study of fully exploiting and reclaiming the remaining execution time of tasks during their executions at runtime and allocates them to subsequent tasks to slow down the processor＇ s frequency via dynamic voltage scaling, while meeting the real-time constraints and task synchronization with as minimum energy consumption as possible. The HDSA avoids frequent voltage switching caused by the frequency inheritance when a task with lower priority blocks the task with higher priority. On that account, the proposed algorithm reduces the scheduling costs efficiently. The experiment results indicate that the HDSA obviously outperforms other algorithms so far in terms of scheduling performance.