Edinburgh Napier University

  Much research has been conducted focusing on improving resource utilization efficiency in data centers in the context of Green Cloud Computing (GCC). While virtualization enables better resource provision and utilization for various computational resources, different approaches are proposed based on virtual machine (VM) optimizations using either server or workload consolidation techniques. Nonetheless, these solutions can only be applied inside the Cloud. In fact, Infrastructure-as-a-Service (IaaS) users can hardly proactively achieve better VM resource utilization efficiency, as they typically have no control over any hypervisor or hardware in any Clouds. The issue gets more critical when workloads on VMs alter dramatically from time to time. This paper presents a novel approach namely Transition and Reallocation Based Green Optimization (TARGO) for such users. Through fully automated and intelligent VM optimization according to customizable optimization rules, TARGO guarantees that VMs or their successors being optimized will always run at their customizable green optimal states regardless how workloads vary. Experiments conducted on Amazon EC2 instances in the EU region show that, even under extreme random workloads, TARGO is still capable of selecting and retaining VM successors which run at an average CPU utilization of 50%-60%.