A scheduling algorithm to reduce the static energy consumption of multiprocessor real-time systems

Energy consumption of real-time embedded systems is a growing concern. It includes both static and dynamic consumption and is now dominated by static consumption as the semiconductor technology moves to deep sub-micron scale. In this paper, we propose a new approach to efficiently use the low-power states of multiprocessor embedded hard real-time systems in order to reduce their static consumption. In a low-power state, the processor is not active and the deeper the low-power state is, the lower is the energy consumption but the higher is the transition delay to come back to the active state. Our approach increases the duration of the idle periods to allow the activation of deeper low-power states. Offline, we use an additional task to model the idle time and we use mixed integer linear programming to reduce its number of preemptions. Online, we extend an existing scheduling algorithm to increase the length of the idle periods. To the best of our knowledge, this is the first optimal multiprocessor scheduling algorithm minimizing static consumption. Simulations show that the energy consumption while processors are idle is dramatically reduced with our solution compared to existing multiprocessor real-time scheduling algorithms.