Evaluating embedded relational databases for large model persistence and query

Abstract

Large models are increasingly used in Model Driven Development. Different studies have proved that XMI (default persistence in Eclipse Modelling Framework) has some limitations when operating with large models. To overcome them, recent approaches have used databases for the persistence of models. EDBM (Embedded DataBase for Models) is an approach for persisting models in an embedded relational database, providing scalable querying mechanism by runtime translation of modellevel queries to SQL. In this paper, we present an evaluation of EDBM in terms of scalability with existing approaches. GraBaTs 2009 case study (models from 8.8MB to 646MB) is used for evaluation. EDBM is 70% faster than the compared approaches to persist XMI GraBats models into databases and executes the GraBats query faster, as well as having a low memory usage. These results indicate that an embedded relational database, combined with an scalable query mechanism provides a promising alternative for persisting and querying large models.