Service reliability modeling of the IT infrastructure of active-active cloud data center

摘要

With the increasing use in different areas, cloud data center has gradually showed its superiority in availability, resource utilization and disaster recovery during the service delivery, compared with traditional data center. According to SLA (Service-Level Agreement), the demand on service reliability and other related indexes are put forward. Despite efforts at fault tolerance and redundancy, the occurrence of failure in data center is still inevitable. Hence, there is a need to model and analyze the service reliability of data center. However, traditional method of reliability modeling is no longer applicable because of the complicated cloud control flows, massive-scale service sharing, and complexity real-word infrastructures. This paper proposes a new approach to model the service reliability of the IT infrastructure of active-active data center, which is a typical form of cloud data center. Firstly, we divide the process of service delivery into two stages - the request stage and execution stage. Then, two models are built for two stages, respectively. For request stage, we use the Queuing Theory and Monte Carlo Method while for execution stage, the Graph Theory and Monte Carlo Method are adopted. Based on the proposed model, the service reliability of data center can be calculated. With the data of a company active-active data center as a case study, we finally demonstrate the applicability and correctness of the model.