Software Defined Network Based Virtual Machine Placement in Cloud Systems

Abstract

As cloud services become mature and provide capabilities for elastic computation and data management, hosting services become increasingly prevalent such as Amazon Web Services, Microsoft Azure, or RedHat OpenStack. Workloads are hosted in these platforms using hypervisors running on physical machines which host virtual machines (VMs) presented to the customer. Software-defined network (SDN) technologies move many of the network functions into the hosts. Accordingly SDN-based VM placements have received significant attention as a way to minimize network traffic between VMs and hosts. Careful placement of VMs can generate a significantly lower amount of network traffic to the hosting provider. Further, it may impact the level of security risk exposed by the VM placement as each VM may have a different level of security risk. In this work, we consider these two conflicting goals by presenting a Software-Defined Network based VM Placement, called SDNVMP, to minimize both network traffic between VMs (and hosts) and security risk by grouping VMs with similar security scrutiny needs. Our results show that the proposed placement algorithm outperforms the baseline counterparts in terms of network traffic overhead and security risk in each VM introduced by such algorithms.