Abstract
In the Era of sharing, several efforts have been launched to construct repositories (referential/warehouses) storing entities used in projects for enterprises (e.g., data, processes, models, APIs, etc.). These repositories augment the business value of the enterprises in terms of reuse, sharing, traceability and analysis. By exploring the literature, we figure out the absence of appropriate warehouses dedicated to functional requirements (\(\mathcal {FR}\)) for the analytical purpose. In a large scale software in the context of global enterprises, involving numerous partners, \(\mathcal {FR}\) may be very heterogeneous in terms of the used vocabularies and formalisms to model them. Other aspects that have to be handled, when constructing a \(\mathcal {FR}\) warehouse are: (a) the management of interdependencies that may exist among \(\mathcal {FR}\) and (b) their scheduling. These aspects complicate the construction of such warehouse. In this paper, we first propose a complete and comprehensive semantic-driven methodology, to design \(\mathcal {FR}\) warehouses. Secondly, all steps of our approach leveraged from the traditional warehouse design are highlighted: (i) the definition of the multidimensional model, (ii) adapting the existing operators of ETL (Extract, Transform, Load) to deal with \(\mathcal {FR}\). ETL uses reasoning capabilities to eliminate the conflictual requirements. (iii) Translating the multidimensional model to its corresponding logical model and (iv) evaluating the performance of the final warehouse. Finally, a proof of concept for our proposal is presented using Oracle DBMS and the vocabulary QB4OLAP proposed by the W3C Government Linked Data Working Group to facilitate the manipulation of semantic warehouses.
