Modelling and simulation for cost optimization and performance analysis of transactional applications in hybrid clouds

Abstract

In the design process of a hybrid cloud with the purpose of cost optimization, finding the optimal distribution between the resources to be deployed in both private and public clouds is a complex task. To help the cloud designer in this task, optimizer and simulation tools can be used. In the research presented in this paper, a state-of-the-art optimizer tool, designed for the optimal allocation of virtual machines in public clouds for cost optimization, is analyzed for its use in hybrid cloud scenarios. In addition, based on the system and workload model implemented by this tool, a simulator is developed and presented. The simulator provides detailed run-time performance information corresponding to the virtual machine allocations generated by the optimizer tool, offering essential analysis capabilities to the cloud designer. A motivating scenario is used to show how the combined utilization of the optimizer and simulation tools can offer fundamental help in the design and configuration process of hybrid clouds. The research questions posed in the motivating scenario are addressed in a set of experiments, which provide meaningful insights into the capabilities of the tools. The tools and all the experimental data carried out in this research have been made publicly available.