关键词: 云计算, 田口方法, 差分进化算法, 多任务调度, 成本和时间模型

Abstract: For the issue that multi-task scheduling and resource allocation problems in cloud computing, an Improved Differential Evolution Algorithm （IDEA） fusing Taguchi method and Differential Evolution Algorithm （DEA） is proposed. The algorithm encodes the tasks by using an Orthogonal （OA） of Taguchi method as a mask mutation operator, produces better offspring by the mutation and crossover process, and establishes a cost and time model in order to find out the Pareto optimal scheduling scheme. A cloud platform environment containing five tasks and five resource is simulated to compare IDEA algorithm with DEA, NSGA-II and other existing algorithms by using average cross rate, distribution distance, maximum width and high-dimensional space ratio as the performance index. Experimental results show that, IDEA algorithm outperforms NSGA-II, DEA and other algorithms in finding Pareto optimal solution for task scheduling and resource allocation. In addition, the optimal scheduling schemes finding out by this algorithm are listed for the targets  with different completion time and task scheduling cost separately which may provide great help to decision-makers.

Key words: cloud computing, Taguchi method, differential evolution algorithm, multi-task scheduling, cost and time model