Abstract
An efficient placement of virtual machines (VMs) in a cloud datacenter is important to maximize the utilization of infrastructure. Most of the existing work maximises the number of VMs to place on a minimum number of physical machines (PMs) to reduce energy consumption. Recently, big data applications become popular which are mostly hosted on cloud datacenters. Big data applications are deployed on multiple VMs and considered data and communication intensive applications. These applications can consume most of the datacenter bandwidth if VMs do not place close to each other. In this paper, we investigate the use of different VM placement methods to decrease the usage of bandwidth in different sizes of datacenters. We implemented and evaluated five different VM placement algorithms. Our comprehensive set of experiments show a significant decrease in bandwidth ranging from 65% to 78% can be achieved using the extended implementations of the knapsack and first fit VM placement methods.