Energy efficient scheduling for real-time embedded systems

Energy management is an important aspect in the design and operation of embedded systems. The processor is central to an embedded system and consumes a significant portion of the total energy consumption. To reduce energy consumption, recent processors support dynamic voltage and frequency scaling which provides the ability to perform an energy-delay tradeoff in the system. Real-time systems have strict timing constraints and voltage scaling decisions must be made wisely while meeting all timing requirements.; We address energy efficient scheduling techniques, based on a variable voltage and frequency processor architecture. We propose novel speed scheduling algorithms that are applicable to tasks with non-preemptive sections and synchronization constraints. We also present the computation of close to optimal speed settings for arbitrary relationship between task deadlines and the task period. Furthermore, we address minimizing both the static and dynamic energy consumption of the system. While energy savings based on voltage scaling come at the cost of increased execution time, this increases the static energy consumption. We propose computation of processor speed settings based on the critical speed---the operating speed that minimizes the total of static and dynamic energy consumption. Furthermore, we combine critical speed slowdown with procrastination scheduling, which extends the duration of idle intervals and minimizes the static energy drain through shutdown. The proposed scheduling techniques support slack reclamation to exploit variations in task execution times for additional energy savings. We address scheduling under both static and dynamic priority systems. Simulations results show the efficiency of the proposed scheduling algorithms.