期刊
JOURNAL OF SUPERCOMPUTING
卷 78, 期 17, 页码 18856-18886出版社
SPRINGER
DOI: 10.1007/s11227-022-04616-y
关键词
Cloud computing; Multiobjective workflow scheduling; List scheduling; Variable neighborhood search; Ant colony optimization
类别
资金
- National Natural Science Foundation of China [61873040, 61973042, 62166027]
- Science and Technology Plan Project of Jiangxi Provincial Education Department [GJJ190959]
- Jiangxi Provincial Natural Science Foundation [20212ACB212004]
Effective workflow scheduling is crucial for improving execution performance in cloud computing. This article introduces a new approach, VMALS, that combines ant colony optimization and list scheduling. Using variable neighborhood search, VMALS achieves better results compared to other algorithms.
Effective workflow scheduling is essential to obtain high execution performance of workflow applications in cloud computing and remains a challenging problem. Due to the commercial nature of clouds, the execution cost of a workflow is a crucial issue for cloud users except for the execution time (makespan). We formulate the cloud workflow scheduling as a multiobjective optimization problem to minimize both execution cost and makespan. A Variable neighborhood search-based Multiobjective Ant colony optimization (ACO)-List Scheduling approach (VMALS) is proposed to address it. In VMALS, the list scheduling is first integrated into the ACO-based multiobjective optimization to consider the effect of different task scheduling sequences on the execution cost and makespan of a workflow. Then, a variable neighborhood search (VNS) is applied to nondominated solutions generated by ACO to approximate the true Pareto front better. Moreover, two novel crossover and mutation-based neighborhood structures are devised to enhance the local search capability of VNS. VMALS is compared with some state-of-the-art algorithms. Experimental results show that VMALS performs better than the comparative algorithms, and the average value of hypervolume metric of VMALS is 3.54-86.18% higher than that of comparative algorithms.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据