Multi-Objective Cloud Task Scheduling optimization based on Evolutionary Multi-factor Algorithm
Zhihua Cui,Tianhao Zhao,Linjie Wu,A.K. Qin,Jianwei Li
DOI: https://doi.org/10.1109/tcc.2023.3315014
IF: 5.697
2023-01-01
IEEE Transactions on Cloud Computing
Abstract:Cloud platforms scheduling resources based on the demand of the tasks submitted by the users, is critical to the cloud provider's interest and customer satisfaction. In this paper, we propose a multi-objective cloud task scheduling algorithm based on an evolutionary multi-factorial optimization algorithm. Firstly, we choose execution time, execution cost, and virtual machines load balancing as the objective functions to construct a multi-objective cloud task scheduling model. Secondly, the multi-factor optimization (MFO) technique is applied to the task scheduling problem, and the task scheduling characteristics are combined with the multi-objective multi-factor optimization (MO-MFO) algorithm to construct an assisted optimization task. Finally, a dynamic adaptive transfer strategy is designed to determine the similarity between tasks according to the degree of overlap of the MFO problem and to control the intensity of knowledge transfer. The results of simulation experiments on the cloud task test dataset show that our method significantly improves scheduling efficiency, compared with other evolutionary algorithms (EAs), the scheduling method simplifies the decomposition of complex problems by a multi-factor approach, while using knowledge transfer to share the convergence direction among sub-populations, which can find the optimal solution interval more quickly and achieve the best results among all objective functions.
computer science, information systems, theory & methods