Challenges in the Automotive Industry: EMF model repositories - scalability vs performance
Eclipse has established itself as the emerging tool platform in the automotive industry. Various companies cooperate in the Automotive Industry Working Group to provide a common Eclipse platform for automotive tooling. A number of these tools will be used to manage the system models of one or more ECUs (Electronic Control Unit).
However, these models can easily become quite large and easily exceed the capacities of 32bit Java virtual machines or standard workplace PCs.
Robert Bosch GmbH has researched various solutions to the challenges of these kinds of models. Model repositories (CDO, EMFStore,...) or light-weight emf-stores (Mongo EMF, CDO Light,..) have been studied, because they promise to provide a solution to scale, share and store emf-based models. When the models don't fit into memory anymore, model repositories are using persistence technologies which are known from the enterprise or web development domain to scale the models.
In addition, they also provide mechanisms to share models among multiple users and make it easy to work on them simultaneously, through fine grained locking & transaction mechanisms. However, system engineering models differ considerably in structure and usage from business data. Queries and validations can easily traverse large parts of a model. That implies that for the development of Eclipse-based tooling for automotive, aerospace or railway, the straight-forward integration of a model repository with of an existing emf-based workbench might have unexpected effects.
The results of the Bosch-funded research, that are being presented in the this talk, on integrating a big emf-based workbench with a model repository from a performance perspective are forwarded to the Automotive Industry Working Group. In its work package on large model support, approaches to address these challenges are being discussed.