Towards a Framework for Reasoning about the Performance of Component Software

The issues concerning the performance of commercial-off-the-shelf (COTS) components have gained much attention in the research community recently. Predicting the performance of a system constructed from components is beneficial for reasons such as bottleneck identification and performance tuning. However, the research status in this area is still relatively immature. Component technologies such as .NET and Enterprise JavaBeans offer assembling platforms but do not provide any insights into the system performance before construction. In this paper, the Performance Characterisation And Composition (P-CAC) framework is used to address the prediction problem. P-CAC introduces component performance specification using the concept of atomic operations that are derived using traditional algorithmic analysis. We refine the connector set proposed by Mehta et al. [10] by incorporating the connectors’ properties as part of the interface specifications for reasoning about its performance. A web of interacting processes which is the consequence of the properties and configurations of the components/connectors, is then used to deduce the performance of the composed system. We discuss these findings and suggest further research directions.