Power Efficient Scheduling Algorithms for Real-time Tasks on Multi-mode Microcontrollers

Abstract

Mobile smart devices are advancing with stronger demands of high energy efficiency and longer battery life. Utilizing energy-efficient microcontroller units in a mobile device for always-on functionalities has proven to be an effective solution. MCUs have the ability to switch between different running modes dynamically enabling them to have outstanding low power performance while performing real-time sensing tasks. Besides hardware optimization, balancing energy efficiency and quality of service on a MCU lies within a well-designed scheduling algorithm. In this paper, we formally define, model and derive a proper scheduling algorithm that guarantees the hard real-time task set's schedulability and minimizes power consumption. Our findings provide an approach of calculating the MCUs resource shutdown schedule, i.e., the shutdown period and the standby time in each period, under Earliest Deadline First and Rate Monotonic scheduler. However, periodically shutting down the MCU may be a simplified way of implementing real-time scheduling on a MCU but not necessarily the optimal approach. Therefore, we further use a simulation to evaluate the performance gap of the periodic shutdown method and the optimal shutdown method with respect to the power savings.