The Impact of Inter-Virtual Machine Traffic on Energy Efficient Virtual Machines Placement

In this work, we investigate the energy efficiency of placing virtual machines (VMs) in geo-distributed data centers taking into account inter-VM traffic in addition to users traffic. The problem of VMs placement is formularized as a mixed integer linear programming (MILP) model with an objective to minimize the network and cloud power consumption taking into consideration cooperation traffic between different VMs and synchronization traffic between replicas of the same VM in addition to the download traffic from VMs to users. The model results show that the number of VMs replicas across geo-distributed clouds is limited by the existence of inter-VM traffic in the core network. The total power consumption can potentially increase by a factor of 39 if inter-VM traffic is not taken into consideration when optimizing the placement of VMs.