A Traffic-Aware Virtual Machine Placement Method for Cloud Data Centers

Abstract

Classical virtual machine (VM) placement approaches, focusing on computing resource requirements, do not consider network traffic as an input for the placement problem, which may result in high network traffic within the core links of the data center. Cost of networking is not negligible in a cloud data center. In order to decrease networking cost, in terms of power and delay, communication patterns of VMs need to be taken into account. We develop an algorithm with low computational complexity, to cluster VMs based on dynamic network traffic data and place clusters into racks, composed of physical servers. Our aim is to decrease the traffic between racks, by putting frequently communicating VMs together into the same rack or as close as possible, while minimizing networking delay based on average communication path length, number of active servers and number of active network elements, such as links and switches. Through numerous simulations, we show that the proposed algorithm provides quick and effective results to improve placement, which can be utilized via VM migrations.