Low-power dynamic scheduling algorithm for real-time multiprocessor systems

Abstract

DVS(Dynamic Voltage Scaling) is one of the important low-power techniques. To reduce power consumption, lowering the voltage of active tasks using the idle time of a processor increases processing time by reducing the frequency. But hard real-time systems under strong time constraints need the powerful multiprocessors which can execute tasks in a fixed time. Therefore the latest embedded systems require high performance as well as minimum power consumption. This paper proposes the GMD(Greedy Myopic-based DVS) algorithm. The GMD algorithm satisfies time constraints of realtime multi-processor systems and low-power design, as applying DVS to the myopic algorithm based on the heuristic function for the dynamic scheduling of tasks have time constraints and resource requirements. While the GMD algorithm yields a performance similar to that of myopic algorithm, it can also reduce the power consumption of multiprocessor systems. ©2008 IEEE.