Composition and tradeoff of non-functional attributes in software systems: research directions

The domain of our work is focused on the study of the quality attributes of component-based systems and service-based systems. The qualities of a component-based system depend on those ones of the single components. The objective of our work is to support the selection of (COTS and in-house) components under the satisfactionof the quality attributes. First, we have introduced an optimization model, that supports the component selection activity. The selection criterion is based on cost minimization of the whole system subject to constraints on reliability and delivery time of the whole system. In addition, for each in-house developed component it suggests the amount of testing to perform in order to achieve the required level of reliability. In order to automatize the selection activity we have introduced CODER (Cost Optimization under DElivery and Reliability constraints), a framework that allows to annotate UML diagrams with cost, reliability and delivery time information and that automatically generates and solves the optimization model. In order to improve the formulation of our model we are working on the reliability expression. We express the reliability of the whole system as function of the reliability of the single components by assuming independence among components. We have relaxed this assumption and we have introduced a path-based reliability model for a Service-Oriented Architecture (SOA) that embeds the "errorpropagation" property. Currently, we are working to introduce a framework that automatizes the process of selection of (COTS and in-house) components in all the phases of a component-based development process. In particular, we have introduced a framework thatsupports the selection of COTS components in the requirements phase. The framework lays on a tool that builds and solves an optimization model, whose solution provides the optimal COTS component selection. The selection criterion is based on cost minimization of the whole system while assuring a certain degree of satisfaction of the system requirements.