A dynamic voltage scaling with single power supply and varying speed factor for multiprocessor system using genetic algorithm

Abstract

With growing of applications of the embedded system technology to mobile systems, energy efficiency is becoming an important issue for designing real time embedded systems. One of the possible techniques to reduce the energy consumption is the Dynamic Voltage Scaling (DVS). DVS utilizes the slack time and adjusts the supply voltage so as to reduce the energy expense. However, how to optimally adjust the supply voltage is a NP hard problem. This paper focuses the combinational optimization problem, namely, the problem of minimizing schedule length with energy consumption constraint and the problem of minimizing energy consumption with schedule length constraint. These problems emphasize the tradeoff between power and performance and are defined such that the power-performance product is optimized by fixing one factor and minimizing the other. We propose the analytical result which gives the variation factor of each power supply which depends on the workload and provides the same power supply while meeting the constraints. We address to the use of genetic algorithm to schedule the tasks and then find the optimal power supplies and determine the schedule length on the multiprocessor system.