Appropriate performance evaluations of commercial Cloud services are crucial and beneficial for both customers and providers to understand the service runtime, while suitable experimental design and analysis would be vital for practical evaluation implementations. However, there seems to be a lack of effective methods for Cloud services performance evaluation. For example, in most of the existing evaluation studies, experimental factors (also called parameters or variables) were considered randomly and intuitively, experimental sample sizes were determined on the fly, and few experimental results were comprehensively analyzed. To address these issues, the authors suggest applying Design of Experiments (DOE) to Cloud services evaluation. To facilitate applying DOE techniques, this paper introduces an experimental factor framework and a set of DOE application scenarios. As such, new evaluators can explore and conveniently adapt our work to their own experiments for performance evaluation of commercial Cloud services.
Boosting Metrics for Cloud Services Evaluation -- The Last Mile of Using Benchmark Suites