Parallel Evolutionary Algorithms for Energy Aware Scheduling

Abstract

Reducing energy consumption is an increasingly important issue in computing and embedded systems. In computing systems, minimizing energy consumption can significantly reduces the amount of energy bills. The demand for computing systems steadily increases and the cost of energy continues to rise. In embedded systems, reducing the use of energy allows to extend the autonomy of these systems. In addition, the reduction of energy decreases greenhouse gas emissions. Therefore, many researches are carried out to develop new methods in order to consume less energy. This chapter gives an overview of the main methods used to reduce the energy consumption in computing and embedded systems. As a use case and to give an example of a method, the chapter describes our new parallel bi-objective hybrid genetic algorithm that takes into account the completion time and the energy consumption. In terms of energy consumption, the obtained results show that our approach outperforms previous scheduling methods by a significant margin. In terms of completion time, the obtained schedules are also shorter than those of other algorithms.

Assessing the Use of Genetic Algorithms to Schedule Independent Tasks Under Power Constraints

“…A parallel bi-objective hybrid genetic algorithm for reducing energy consumption and makespan in computing systems is presented in [13]. Each chromosome represents a possible solution, and each gene assigns one task to a processor and sets its voltage.…”

“…Lei et al [14] propose a multi-objective co-evolutionary algorithm for scheduling tasks on data centers partially powered by renewable energy. The chromosome encoding is similar to the one in [13]. The processor and voltage parts of genes are the ones who change through a chromosome by mutation and crossover, while the order of tasks remains fixed according to their index.…”

“…Both works ( [13] and [14]) use DVFS to set the voltage of a processor, while the model considered here assumes that different tasks have different fixed power consumptions. Also they only partially supply the data center with renewable energy and have recourse to brown energy sources.…”

Energy efficient genetic‐based schedulers in computational grids

“…Kessaci et al in [21] present two versions of multi-objective parallel genetic algorithm hybridized with energyconscious scheduling heuristics. The solution presented in [21] is dedicated to general computing and embedded systems. The authors used parallel applications represented by a directed acyclic graph that are mapped into multi-processor machines.…”

“…Although the single-population GAs are very simple for the implementation and adaptation to the various scheduling criteria, the multi-objective genetic algorithm (MOGA) [20] seems to be the key solution to tackle the complexity of the multi-criteria grid scheduling process. Kessaci et al in [21] present two versions of multi-objective parallel genetic algorithm hybridized with energyconscious scheduling heuristics. In this algorithm, the GA engine is based on the concepts of island GA and multi-start GA models.…”