Efficient evaluation of alternatives for assembly of services

Component based software engineering (CBSE) provides rapid development using well-tested components with established properties. Performance and other nonfunctional properties can also be analyzed by building models from sub-models, calibrated for the components. Further there can be many choices of components to build-systems, which can provide alternatives. The choice can be governed by goal functions which evaluate the predicted performance. This paper describes a systematic approach to find the feasible combinations of alternatives, and to rank them based on predicted performance. It extends the CBML (component based modeling language) for defining components in layered queuing models for software performance.