Optimisation of Model Management Programs Using Automated Program Rewriting

Over the last few years, the use of Model Driven Engineering (MDE) in industrial applications has been increasing rapidly. For the use of MDE-based applications at a larger scale, they should scale well in terms of the size of the model, execution time, and memory consumption. Adapting the use of MDE for extensive software landscapes, where the underlying models grow large in size poses various scalability challenges. When model management tools run on pay-as-you-go cloud-based resources, inefficiency and limited scalability incur substantial costs. Hence, there is vested interest from vendors of cloud-based MDE solutions for efficient and scalable model management tools. There are specific high-level languages to develop model management programs tailored for the specific tasks they target. This work aims to improve the performance of certain types of model management programs through static analysis. An approach is proposed for optimising model management tasks,
particularly model validation, model-to-model transformation and model comparison over large-scale models. The proposed approach leverages static analysis and automated program rewriting techniques to optimise model management programs over large-scale EMF-based models. This optimisation approach aims to bring efficiency in terms of execution time and memory footprint so that developers can still express model management programs in high-level language and execute these programs efficiently. The program is automatically rewritten to an optimised version (where possible). The optimised program is semantically equivalent to the original program but faster and more efficient to execute. The experiments of this study have shown a significant performance gain in
execution time and memory footprint.