Application of parallel metaheuristics to an execution time-power consumption bi-objective problem
1 : University of Murcia
* : Corresponding author
This work studies a bi-objective problem in which the execution time and the energy consumption of an algorithm are optimized. A master-slave scheme in which a master sends identical tasks to slave processes is considered. The processes are assigned to processors in a heterogeneous cluster, with computation, communication and energy heterogeneity. Models of the execution time and the energy consumption are used, and the mappings giving time-energy pairs in the Pareto front are searched for. Genetic metaheuristics for the problem are developed, with sequential, shared-memory, message-passing and hybrid versions. Satisfactory results are obtained both in the goodness of the solutions and the performance of the parallel versions.