Service reliability modeling and evaluation of active-active cloud data center based on the IT infrastructure

Abstract As cloud data center has caught the eye of the information-intensive society since its birth, it keeps on a flourishing development due to its advantages such as high availability and resource utilization, rapid elasticity and disaster recovery. However, as a complex system, it means the centralization of failures and risks, which exerts great influence on the service that cloud data center provides to customers. Thus, the service reliability of the cloud data center is always a key concern. In order to evaluate and further improve the service reliability of the cloud data center, modeling analysis is absolutely necessary but difficult to apply because of the complicated cloud control flows, massive-scale service sharing and complex real-world infrastructures. Focused on the active-active data center, which is a typical mode of the cloud data center, a methodology of the service reliability modeling and analysis based on IT infrastructure is proposed. In this methodology, the queuing theory and graph theory are used to formulate the service reliability model, but the Monte Carlo simulation is used for statistical evaluation. During the modeling, the operational process of the active-active data center is divided into two parts—the request stage and execution stage, which are modeled respectively. Then, the modeling and evaluation approaches are applied in a use case, which verifies the applicability and creditability of our approach. Meanwhile, by sensitivity analysis considering the variation of uncertain or key factors both internal and external, several parameters are identified as high-sensitive factors, which can enlighten service providers on service reliability improvement.