Quantitative management of software production using object-oriented models: Dissertation

Managing the process of constructing software intensive products is still a major problem, despite the introduction of new methods, techniques, and tools. We can tackle this problem with quantitative management of the software process. Since we do not have concrete objects to measure and control, we need to have rigorous models of the knowledge and the various elements in software engineering. Given an approach to quantitative management and learning in a software organization, we need to support it by building automated, integrated models of the software process. We present a model-based approach for representing software engineering knowledge. The model- based approach has been used to develop a method and an intelligent support system (Prospex) for assisting embedded computer software analysis and design. This system can be used to measure the quality and correctness of design documents, and thus help in improving the early phases of the software development process. The object-oriented mechanisms have been extended with predefined relationships, dynamic hierarchies and selective inheritance in a prototype environment called ES-TAME. They are used to demonstrate a framework for improving software production processes with quantitative methods. ES-TAME supports two types of models. GQM models define goals which are refined into quantitative questions and metrics for the software process and products. Software engineering models are used to represent various kinds of descriptive knowledge of software processes, products, design methods, resource models, etc. Both models are represented with extended object-oriented models which are interconnected into an environment which can be used in a goal oriented quantitative method for controlling and improving software processes. The two prototype systems demonstrate the feasibility of automating and integrating extended object-oriented models for supporting software production, as well as controlling and improving software processes. The underlying logical models serve as a foundation for intelligent tools and mechanisms for representing and organizing knowledge of various software engineering elements. The model-based approach provides means for extendibility, flexibility and techniques for viewing the models from multiple perspectives.