Modeling and Solution Algorithm of Virtual Machines Optimization Provision Problem for Application Deployment in Public Cloud

Fanchao Meng, Qingran Ji, Dianhui Chu, Xuequan Zhou

Abstract: At present, when applications are migrated to the public cloud, the decision-making of virtual machine provision for applications deployment mainly depends on the personal experience of service providers, which leads to the high rental cost of cloud resource. In this paper, we propose a virtual machines optimization provision problem (VMOPP) for applications deployment in public cloud. The objective of VMOPP is to minimize the total sum of costs of used virtual machine instances and communication costs between virtual machine instances. The VMOPP is modeled as an integer programming problem. A variable neighborhood search based on hierarchical clustering algorithm (VNSBHCA) is proposed to solve this problem. To evaluate the performance of the proposed VNSBHCA for VMOPP, the hybrid simulated annealing genetic algorithm (HSAGA) and commercial solver CPLEX were also applied to VMOPP. The performance of proposed VNSBHCA, HSAGA and CPLEX was evaluated using 20 generated test instances. Experimental results show VNSBHCA can search better solutions than HSAGA and CPLEX.