Cloud Computing Task Scheduling Based on Improved Differential Evolution Algorithm

Abstract

In this paper, we propose an improved discrete differential evolution algorithm to solve the problem of cloud computing task scheduling, which can greatly improve the computational efficiency of task scheduling in cloud computing. The purpose of this improved algorithm is to obtain the shortest completion time of a total task and get a better load balance. In the process of population initialization, each individual of the population is coded through natural number, and the crossover constant changes with iteration times, which is used to make a balance between global search and local search. After mutation and crossover operation, a new selection mechanism will be adopted to preserve those excellent individuals among mutation individual, target individual and trial individual. Numerical experiment runs in Cloudsim platform (a computer simulation platform used to stimulate cloud computing environment). Our experimental results show that compared with standard DE algorithm, the improved differential evolution algorithm has a better performance on convergence and a good ability to jump out from local optimum, which means the algorithm can get a better scheduling result for cloud computing. In the experiment, we set the numbers of subtasks as 50, 200 and 500 and the completion time is reduced by 0.83s, 10.94s and 84.13s respectively.