A Methodology for Automating the Cloud Data Center Availability Assessment

Abstract

Organizations worldwide are rapidly migrating their IT services and infrastructure to cloud data centers in order to reduce costs and guarantee high availability, scalability, and security. Considering that service downtime translates into major financial losses, new mechanisms need to be developed to assess the availability of cloud data center dynamically. However, data center availability analysis remains a complex task and one that is prone to human error due the large number of components and their interconnections. In this work we propose a methodology for acquiring information about a data center infrastructure and, automatically, generating computational models to assess its availability. We make use of the Redfish standard to acquire information about the data center infrastructure, the main standard for data center management. To demonstrate the applicability of our proposal, we conduct a study to analyze availability and failure costs of an application hosted in a cloud and we compare different scenarios with redundant servers according to the TIA-942 data center standard. Results show that a lower tier level with redundant servers, in some cases, is more suitable (more available and less costly) than higher tier levels without redundant servers hosting a cloud application.