Model-driven software development is an increasingly popular development paradigm. In this method, developers start with an abstract and high level model of the system, and from that they make an executable application through some model transformation steps.
Model transformations are one of the enabling technologies for model-driven development. It aims to carry out automated translation within and between modeling languages. With the widespread of industry driven standards (like AUTOSAR and AADL) based on the principles of model based development, the need to handle transformation between millions of model elements has become a key question. In the industry the information are stored in relational databases. So it is a handy solution to store models in relational databases. However modeltransformation over databases suffer from performance issues: they are way slower than the modeltransformation frameworks. Instead of using a general relational database, in-memory databases can be a good solution for this problem.
In model transformation frameworks model elements are stored as heavy (Java) objects with lots of references and collections. So large models cannot be stored and processed because of memory issues.
So achieving model transformation over relational databases is increasingly important. The goal of my thesis is developing an efficient tool, that achieves model transformation over in-memory databases. The mapping of metamodels describing the syntax of models, the mapping of graph transformation rules including the graph patterns and manipulation steps, and importing instance models into relational databases are the objectives of the thesis.
The functional and performance evaluation of the implemented software component is very important from the viewpoint of further usage. The generated scripts were tested by different benchmarks and test cases.