Conservative and traceable executions of heterogeneous model management workflows

One challenge of developing large scale systems is knowing how artefacts are interrelated across tools and languages, especially when traceability is mandated e.g., by certifying authorities. Another challenge is the interoperability of all required tools to allow the software to be built, tested, and deployed efficiently as it evolves. Build systems have grown in popularity as they facilitate these activities. To cope with the complexities of the development process, engineers can adopt model-driven practices that allow them to raise the system abstraction level by modelling its domain, therefore, reducing the accidental complexity that comes from e.g., writing boilerplate code. However, model-driven practices come with challenges such as integrating heterogeneous model management tasks e.g., validation, and modelling technologies e.g., Simulink (a proprietary modelling environment for dynamic systems). While there are tools that support the execution of model-driven workflows, some support only specific modelling technologies, lack the generation of traceability information, or do not offer the cutting-edge features of build systems like conservative executions i.e., where only tasks affected by changes to resources are executed. In this work we propose ModelFlow, a workflow language and interpreter able to specify and execute model management workflows conservatively and produce traceability information as a side product. In addition, ModelFlow reduces the overhead of model loading and disposal operations by allowing model management tasks to share already loaded models during the workflow execution. Our evaluation shows that ModelFlow can perform conservative executions which can improve the performance times in some scenarios. ModelFlow is designed to support the execution of model management tasks targeting various modelling frameworks and can be used in conjunction with models from heterogeneous technologies. In addition to EMF models, ModelFlow can also handle Simulink models through a driver developed in the context of this thesis which was used to support one case study.