Integrated Optimization of IT Service Performance and Availability Using Performability Prediction Models

Optimizing the performance and availability of an IT service in the design stage are typically considered as independent tasks. However, since both aspects are related to one another, these activities could be combined by applying performability models, in which both the performance and the availability of a service can be more accurately predicted. In this paper, a design optimization problem for IT services is defined and applied in two scenarios, one of which considers a mechanism in which redundant components can be used both for failover as well as handling overload situations. Results show that including such aspects affecting both availability and performance in prediction models can lead to more cost-effective service designs. Thus, performability prediction models are one opportunity to combine performance and availability management for IT services.