Partitioning and Placing Virtual Machine Clusters on Cloud Environment

Abstract

In a virtualized cloud environment, a cloud scheduler needs to place a set of virtual machines on the compute pool consisting of physical machines. Prior studies usually focus on the issue of placing virtual machines that are assumed to be isolated to each other. This paper addresses the problem of deploying virtual machine clusters on the compute pool, in which a virtual machine cluster refers to a set of virtual machines hosting a virtualized distributed system/application. In the case that the virtualized distributed system/application is too large to fit in a single physical machine, the virtual machine cluster has to be partitioned into several subsets and to be placed on several physical machines. Without considering the relationship among VMs in a virtual machine cluster, two VMs that frequently communicate with each other can be placed far away, and consequently lead to poor performance. To this end, we propose a novel placement approach for virtual machine clusters that tries to minimize the Inter-VMs network traffic. The simulation results indicate that the proposed approach consumes the smallest physical network bandwidth while comparing with several existing virtual machine placement algorithms.