Early Performance Evaluation of Data Warehouse Systems: From UML to LQN models

Abstract: If the performance of a Data Warehouse System is determined to be unacceptable, at the time of �acceptance testing� it can result in very expensive redesign and consequent delayed delivery or, in the worst case, complete non-use of the system! There is clearly a need for tools and techniques that enable performance analysis of designs to be done easily and reliably throughout the development process of Data warehouse systems. In this paper we demonstrate the derivation of Layered Queuing Network (LQN) Performance Models from a set of UML diagrams and an algorithm for deriving LQN model. LQN model is a very useful tool to analyse the performance of a system from abstract model so that the developer of Data warehouse systems is able to understand performance effects of various design decisions starting at early stages when changes are easy and less expensive. Keywords: Data warehouse Systems; Software Performance Prediction; UML; Queuing models.