Application of deterministic, stochastic, and hybrid methods for cloud provider selection

Abstract

Abstract Cloud Computing popularization inspired the emergence of many new cloud service providers. The significant number of cloud providers available drives users to complex or even impractical choice of the most suitable one to satisfy his needs without automation. The Cloud Provider Selection (CPS) problem addresses that choice. Hence, this work presents a general approach for solving the CPS problem using as selection criteria performance indicators compliant with the Cloud Service Measurement Initiative Consortium - Service Measurement Index framework (CSMIC-SMI). To accomplish that, deterministic (CPS-Matching and CPS-DEA), stochastic (Evolutionary Algorithms: CPS-GA, CPS-BDE, and CPS-DDE), and hybrid (Matching-GA, Matching-BDE, and Matching-DDE) selection optimization methods are developed and employed. The evaluation uses a synthetic database created from several real cloud provider indicator values in experiments comprising scenarios with different user needs and several cloud providers indicating that the proposed approach is appropriate for solving the cloud provider selection problem, showing promising results for a large-scale application. Particularly, comparing which approach chooses the most appropriate cloud provider the better, the hybrid one presents better results, achieving the best average hit percentage, dealing with simple and multi-cloud user requests.

Keywords

• Cloud provider selection problem
• Cloud provider ranking
• Evolutionary computing
• Natural computing
• Performance indicator
• Hybrid methods