Semantic Data Warehouse Design: From ETL to Deployment à la Carte

In last decades, semantic databases (\(\mathcal{S}\mathcal{D}\mathcal{B}\)) emerge and become operational databases, since the major vendors provide semantic supports in their products. This is mainly due to the spectacular development of ontologies in several domains like E-commerce, Engineering, Medicine, etc. Contrary to a traditional database, where its tuples are stored in a relational (table) layout, a \(\mathcal{S}\mathcal{D}\mathcal{B}\) stores independently ontology and its instances in one of the three main storage layouts (horizontal, vertical, binary). Based on this situation, \(\mathcal{S}\mathcal{D}\mathcal{B}\) become serious candidates for business intelligence projects built around the Data Warehouse (\(\mathcal{D}\mathcal{W}\)) technology. The important steps of the \(\mathcal{D}\mathcal{W}\) development life-cycle (user requirement analysis, conceptual design, logical design, ETL, physical design) are usually dealt in isolation way. This is mainly due to the complexity of each phase. Actually, the \(\mathcal{D}\mathcal{W}\) technology is quite mature for the traditional data sources. As a consequence, leveraging its steps to deal with semantic \(\mathcal{D}\mathcal{W}\) becomes a necessity. In this paper, we propose a methodology covering the most important steps of life-cycle of semantic \(\mathcal{D}\mathcal{W}\). Firstly, a mathematical formalization of ontologies, \(\mathcal{S}\mathcal{D}\mathcal{B}\) and semantic \(\mathcal{D}\mathcal{W}\) is given. User requirements are expressed on the ontological level by the means of the goal oriented paradigm. Secondly, the ETL process is expressed on the ontological level, independently of any implementation constraint. Thirdly, different deployment solutions according to the storage layouts are proposed and implemented using the data access object design patterns. Finally, a prototype validating our proposal using the Lehigh University Benchmark ontology is given.