Interdisciplinary Variability Modeling and Performance Analysis for Long-Living Software Systems
Matthias Kowal · Digitale Bibliothek Braunschweig (Verbundzentrale Göttingen (VZG)) · 2018
In this day and age, automation systems have to deal with differing customer needs, environmental requirements and multiple application contexts. Automation systems have to be variable enough to satisfy all of these demands. The development and maintenance of such highly-customizable systems is a challenging task and becomes increasingly more difficult considering multiple involved engineering disciplines and long lifetimes, which is characteristic for industrial systems of the automation domain. Software product line engineering provides developers with fundamental concepts to manage the variability of such systems. However, these concepts are not established in the domain of automation systems. In addition, the involvement of multiple engineering disciplines poses a threat to existing SPL techniques. This thesis contributes novel approaches to improve the development and maintenance of software-intensive automation product lines. In total, three major contributions are made, spanning across the complete design phase of an automation system. (1) The feature modeling process is improved by detecting hidden dependencies between interrelated feature models from separate engineering disciplines. Furthermore, hidden dependencies and occurring defects in the feature models are explained in a user-friendly manner. (2) A model-driven development approach is introduced consisting of UML models, which are extended with delta modeling to manage variability in the automation product line. The models encompass information that is needed to automatically derive and analyze a performance model. (3) Subsequently, an efficient family-product-based performance analysis is proposed for the previously derived UML models that is vastly superior compared to common product-based approaches. All of these techniques have been evaluated using multiple case studies, with one being a real-world automation system.