Cooperative IaaS Resource Management: Policy and Simulation Framework

Resource management is one of the main services that all cloud models must ensure. This objective can be done either by the customer (using an application manager) or by the provider (with a server consolidation system). A cooperative resource management policy exploits elastic VMs and knowledge sharing between the customer and the provider in order to support splitting and merging virtual machines. "Splitting" VMs will help in optimizing resource usage while "merging" collocated VMs (ones that are placed on the same physical server) reduces performance and memory overhead. This combined policy aims at saving cost for the two actors and improve performance for the customer. A set of experiments to evaluate the effectiveness of the cooperative resource management policy was performed using a small, private IaaS. However, the verification of this policy must be evaluated in a larger scope. In this paper, we describe the cooperation policy, our cloud resource management simulation framework, and then demonstrate our large scale experiment with our proposed cooperative resource management policy using the framework. The evaluation shows that our policy outperforms uncoordinated resource management in traditional IaaS with better virtualized and physical resource usage.