Metamodel evolution and impact analysis on the transformations – the Gaspard framework case study.

Supervisor: Anne Etien

Project Team: DaRT

Research context:

The DaRT project works on the embedded system design using a Model Driven Engineering approach. In the more precise context of System on Chip (SoC), the Gaspard framework developed by the team enables the design, the simulation, the test and the generation of applications mapped onto material architectures. Gaspard2 is a unified environment dedicated to the SoC co-design:

  • modeling the applications and the material architectures at a high abstraction level (UML+MARTE profile)
  • mapping application onto material architectures (UML+MARTE profile)
  • deploying elementary components (UML)
  • model transformation to automatically target different plateforms.

The Gaspard skeleton is composed of several metamodels that correspond to various descriptions of the same system, possibly at different abstraction levels. Transformations allow the translation from one metamodel to the other and the code generation in various languages depending of the fixed objective.

Awaited work:

In the current version of the Gaspard environment, the application principally relies on the “Structured Component” concept. Each component represents a task that can possibly be composed of other tasks. The design of intensive signal application usually relies on task dependencies defining an ordonnancing graph. We would like to modify the way the applications are designed in order to be in concordance with the habits of the final users, by designing the task with activity and no more with component.

Such a modification in the application design has several repercussions. It is for example necessary to modify the MARTE profile, then the MARTE metamodel. Insofar as the “activity” paradigm can be kept almost until the code, all the other metamodels in the transformation chains need to be modified. And the transformations also.

The evolution of the MARTE metamodel will be used as an example, support to define a more generic proposition. The internship will be composed of two parts:

  • In a first part, the student will propose a new version of the MARTE metamodel and will make adequately evolve the transformations.
  • In a second part, the student will propose more generic techniques to impact on the transformations the metamodel evolutions.

For this second part, the student can focus on some aspects of the metamodel evolution such as renaming or adding. Some works in connexe fields defining typologies of evolution operators, as well as works in MDE on model-metamodel coevolution would be helpful.

Different alternatives will be analyzed: the propagation of the evolution to all the metamodels in the chain or the evolution of only some metamodels. It clearly appears that the impact on the transformation is not the same in these two cases. Prolonging the works performed during this internship in the context of a PhD thesis is possible and will be studied later. The pursue in thesis will for example concern the automation of the transformation adaptations consequently to the metamodel evolutions taking into account the two cases and all the operators.

Required skills:

The selected student shuld have good skills in the Model Driven Engineering technologies. An experiment of their manipulation in the context of Eclipse modeling tools (EMF, QVTO, etc.) would be very useful. Minimally, the student should easily program with Java.

Bibliography
Abdoulaye Gamatié, Sébastien Le Beux, Éric Piel, Anne Etien, Rabie Ben Atitallah, Philippe Marquet, and Jean-Luc Dekeyser. A model driven design framework for high performance embedded systems. Research Report 6614, INRIA, August 2008.

Wachsmuth, G.: Metamodel adaptation and model co-adaptation. In: Ernst, E. (ed.) ECOOP 2007. LNCS, vol. 4609, pp. 600–624. Springer, Heidelberg (2007)

Herrmannsdoerfer, M., Benz, S., Juergens, E.: Automatability of coupled evolution of metamodels and models in practice. In: Czarnecki, K., Ober, I., Bruel, J.-M., Uhl, A., Völter, M. (eds.) MODELS 2008. LNCS, vol. 5301, pp. 645–659. Springer, Heidelberg (2008)

Contact:
Anne Etien
+33 3 59 57 79 07
Anne.Etien@lifl.fr

Localization:
Project Team DaRT
INRIA Lille Nord Europe
Parc Scientifique de la Haute Borne
Park Plaza - Bât A - 40 avenue Halley
59650 Villeneuve d'Ascq - FRANCE