Availability Analysis on Virtualized Two-Node Cluster System: Ratio of Restoration Rate and Failure Rate

Abstract

Worldwide, businesses continually increase their dependence on IT systems for routine business processes. The business processes which directly rely on information systems and the supporting IT infrastructure often require high levels of availability and recovery in the case of planned and unplanned outage. High availability has achieved by host per host redundancy, a highly expensive method with hardware and human costs. Virtualization technologies promise cost reduction through resource consolidation. By combining virtualization and HA clustering, it is possible to benefit from increased manageability and saving from server consolidation through virtualization without decreasing uptime of critical services. Using analytical modeling, we analyze multiple design choices when dual physical servers are used to host multiple virtual machines. We use Markov decision process when we are concerned about optimal decision at any arbitrary time. Numerical examples are presented to illustrate the applicability of the model.