Multithreaded power consumption scheduler based on a genetic algorithm

Abstract

This paper presents a multithreaded power consumption scheduler and measures its performance, aiming at reducing peak load in a scheduling unit. Based on the observation that the same genetic algorithm leads to a different solution for a different initial population, the proposed scheduler makes each thread generate its own initial population and independently run genetic iterations for a better solution. Judging from the performance measurement result obtained from a prototype implementation, multithreaded version can reduce the peak load even with small population size without loss of accuracy. After all, the threaded scheduler improves the computation speed, which is inherently dependent on the population size of a genetic scheduler mainly consist of sorting and selection procedures. © 2011 Springer-Verlag.