Parallel Schedule of Live Migrations for Virtual Machine Placements

Cloud based services play a major role in collaborative computing on the Internet across the globe. Managing the computing and network resources to serve dynamic demands of the tenants of an infrastructure-as-a-service (IaaS) cloud is a major challenge for the cloud service provider (CSP). Accommodating the custom resource requirements often result in the migration of virtual machines (VMs) already executing on a server. Live VM migrations, which relocates a VM without halting, helps the CSP to achieve this objective. Identifying the best set of parallelized migrations which discovers the resources to meet tenants' demands would be highly beneficial to the CSP. This paper proposes a simulated annealing algorithm to find a set of migrations which can be scheduled in parallel so as to complete all the live migrations in minimal time. Completion of the migrations as per the schedule can aid in timely deployment of new VMs. Experimental evaluation shows that the proposed optimal scheduling can significantly speed up the process of resource provisioning compared to the existing techniques and a serial migration schedule.