Bounding the cost of virtual machine migrations for resource allocation in cloud data centers

Abstract

Allocating resources economically to virtual data centers is an important concern for cloud service providers while serving a new virtual data center deployment request. The cloud service providers may want to optimally place these topology adherent virtual data centers on their physical data centers to increase the resource utilization and the revenue. However, multiple provisioning, scaling, and de-provisioning of several virtual data centers of various sizes and topologies leave the cloud data center fragmented in terms of the residual server and network resources. Migrating the existing virtual machines and virtual network, if required, to accommodate a new virtual data center can increase the probability of acceptance and hence, the quality of experience of the customers. However, the migrations are costly and hence should be bounded to increase the revenue. In this paper, we propose a model to find a minimum cost virtual machine migration pattern to accommodate a new virtual data center. The objective is to limit the cost of migration so that the cloud service provider is benefited in terms of revenue. We propose a greedy and a meta-heuristic algorithm to solve the problem. Experimental results show that the proposed technique can reduce the average migration time upto 20% and the penalty for corresponding service level agreement violation by 200%.