Grant: Collaborative Research and Development Grants

Capacity engineering is concerned with selecting the most appropriate configurations for deploying an application while satisfying performance requirements and financial constraints. When deploying enterprise applications, customers must determine whether their current deployment infrastructure, is over-engineered (then they can reduce deployment costs) or under-engineered (then they can invest more to improve the user's experience). However, customers are given little help when conducting capacity analysis; instead customers often uncover limitations through critical crashes and service degradations of already deployed applications. Capacity analysis is an important yet costly and labor intensive task that is given a cursory focus at the end of a project. A capacity analysis for a given software application requires the execution of a large number of experiments, requiring a lengthy process, on a variety of hardware configurations, requiring a costly lab setup. This research project aims to minimize the cost and time associated with capacity engineering efforts by reducing the number of experiments that must conducted and by reducing the number of hardware configurations that should be explored in the experiments. With fewer experiments, capacity analysis activities can be performed within days instead of weeks. With a reduced number of configurations, capacity labs are easier to manage and cheaper to run, due to lower hardware costs and cooling requirements. The results of this research will help companies, like RIM, create and evolve tools, such as capacity calculators, with minimal effort and cost. These tools will assist customers of enterprise applications in their capacity analysis activities and will lead to higher satisfaction for users. The research will train 10 HQP in capacity engineering, an important area of software engineering, for which there is a clear and documented worldwide lack of HQP.