Fine Tuning of Rank Based VM Placement and Scheduling Strategies on Opennebula Based Cloud

Abstract

The HPC Clouds are good option over deploying actual physical infrastructure. This helps in running applications efficiently and in cost effective manner as applications leverage resources in pay as you go manner. But HPC clouds suffer performance issues due to Hypervisor layer. This work addresses the issue by coming up with a VM placement strategy considering the intensity of the applications and also resources available in the host machines and avoid performance degradation of parallel running applications. This placement strategy identifies the and ranks the available host machines and places the maximum possible VMs in highest ranking nodes. This avoids communication over the network since the VMs use shared memory for communication.