A Multi-Workflow Scheduling Approach With Explicit Evolutionary Multi-Objective Multi-Task Optimization Algorithm in Cloud Environment

Abstract

Workflow tasks in the cloud environment are the abstraction and decomposition of large-scale and complex tasks in real-world scenarios, so cloud workflow scheduling problems have important research significance. However, most of the existing cloud workflow scheduling schemes are aimed at a single workflow, and do not make reasonable use of the commonality or complementary knowledge between similar tasks. Moreover, most cloud workflow scheduling models mainly focus on a few objectives such as time or cost, which is not comprehensive enough. Therefore, this paper first proposes a multi-objective cloud workflow scheduling model, which solves the maximum completion time, execution cost and energy consumption as three objectives during task execution. Secondly, to efficiently handle multiple similar cloud workflow scheduling tasks at the same time, this paper treats various cloud workflow scheduling issues as distinct tasks, establishes a multi-task cloud workflow scheduling framework that aims for the same goal while accommodating workflows of differing scales, and a multi-objective evolutionary multi-task optimization algorithm based on elite selection (MOEMT-ES) is designed to solve the above scheduling model. Finally, through algorithm comparison experiments on the CEC2017 evolutionary multi-task optimization competition benchmark problem and multi-workflow test problem, MOEMT-ES shows superior competitiveness.