Evaluating Evolutionary Software Systems

non-functional requirements (NFRs) of software-intensive systems that are under continuous evolution should be evaluated during early development phases in order to be able to improve those systems and achieve ‘time-tomarket’. However, current evaluations are often done during late stages, like coding and testing. In this paper we propose an approach to evaluate NFRs earlier. The requirements for this approach are the use of flexible and reusable quality models, which can deal with little data, that are transparent and measurement-based. Our approach, called Prometheus, is a way of modeling NFRs that should cope with those requirements. Prometheus applies the quality modeling concept from the SQUID approach, the probability concept of Bayesian Belief Nets (BBNs) and the specification concepts of the Goal Question Metric (GQM) approach.