Towards scalable model indexing

• Main Author: Barmpis, Konstantinos
• Other Authors: Kolovos, Dimitrios S.
• Published: University of York 2016
• Subjects: 005.1
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.696076

Model-Driven Engineering (MDE) is a software engineering discipline promoting models as first-class artefacts of the software lifecycle. It offers increased productivity, consistency, maintainability and reuse by using these models to generate other necessary products, such as program code or documentation. As such, persisting, accessing, manipulating, transforming and querying such models needs to be efficient, for maintaining the various benefits MDE can offer. Scalability is often identified to be a bottleneck for potential adapters of MDE, as large-scale models need to be handled seamlessly, without causing disproportionate losses in performance or limiting the ability of multiple stakeholders to work simultaneously on the same collection of large models. This work identifies the primary scalability concerns of MDE and tackles those related to the querying of large collections of models in collaborative modeling environments; it presents a novel approach whereby information contained in such models can be efficiently retrieved, orthogonally to the formats in which models are persisted. This approach, coined model indexing leverages the use of file-based version control systems for storing models, while allowing developers to efficiently query models without needing to retrieve them from remote locations or load them into memory beforehand. Empirical evidence gathered during the course of the research project is then detailed, which provides confidence that such novel tools and technologies can mitigate these specific scalability concerns; the results obtained are promising, offering large improvements in the execution time of certain classes of queries, which can be further optimized by use of caching and database indexing techniques. The architecture of the approach is also empirically validated, by virtue of integration with various state-of-the-art modeling and model management tools, and so is the correctness of the various algorithms used in this approach.