Metamodel driven model migration

• Authors: Jonathan Mark Sprinkle;Gabor Karsai
• Venue: Metamodel driven model migration
• Year: 2003

Model-integrated computing (MIC) is a methodology for the development of computer-based systems (CBSs). MIC predicates that models are developed according to a particular metamodel (that is, a language representation of the CBS's actual domain). MIC is the integration of models with a way to extract a “meaning” from those models. Using metamodeling, it is possible to model the domain of the CBS and generate a domain-specific modeling environment (DSME) whose ontology is the types of components in the CBS. With this DSME, models of the CBS are constructed and expressed with the ontology of the domain, and then interpreted to produce an executable model. The true value of domain-specific modeling is found not in the DSME, but the models that are created in that DSME. Changes to the physical system can be modeled, and the resulting executable model then is a working version of the CBS. Unfortunately, if the model of the domain—or metamodel—is changed, all models that were defined using that metamodel may require maintenance to have the semantics that represent the CBS correctly. Without ensuring the correctness of the domain models after a change to the domain, the true value of the DSME will be lost. The only way to use instance models based on the original metamodel is to migrate them for use in the modified metamodel. This problem is known as the Model Migration (MM) problem. Current state of the art for Model Migration heavily depends on low-level implementation, and treats the domain models like data files rather than domain-specific models. There are benefits to performing model migration using domain-specific concepts such as mapping objects from one domain to another, and leveraging the existing metamodel structure for designing these mappings. This dissertation extends the state of the art by implementing such a metamodel based model migration language, which is a modeling language in its own right. Furthermore, a model migration language is dependent on two components: the meta-metamodel, and the graph-rewriting language used to perform the actual to reveal a framework that any meta-metamodel and graph-rewriting language would use to perform model migration. This dissertation gives form and function to such a framework, and describes how it can be specialized to yield a particular instance of a framework: a domain evolution tool.