Empirical Investigations of Model Size, Complexity and Effort in a Large Scale, Distributed Model Driven Development Process

Model driven development (MDD) is a software engineering practice that is gaining in popularity. We aim to investigate to what extend it is effective. There is a lack of empirical data to verify the pay-offs of employing MDD tools and techniques. In order to increase the knowledge we have of the impact of MDD in large scale industrial projects, we investigate the project characteristics of a large software development project in which MDD is used in a pure form. This study focuses on analyzing model size and complexity and metrics related to model quality and effort. Furthermore, project team members were asked to elaborate on their views on the impact of using MDD. Our findings include that larger models are more complex, contain more diagrams, are changed more often and worked on longer but do not necessarily contain more defects. However, models that are changed often do contain more defects. Benefits mentioned by team members were an increase in productivity, benefits from a consistent implementation and their perception of improvement of overall quality. Also, a reduction in complexity was attributed to the use of MDD techniques. We could confirm the perceived increase in the quality of the product in that the average amount of defects found is significantly lower than in similar size projects in which MDD was not employed.