On Using Requirements Throughout the Life Cycle of Data Repository

Abstract

Requirements engineering aims at providing a requirement specification with some nice properties such as completeness or accuracy. In the lifecycle of a Data Repository ($\mathcal{D}\mathcal{R}$), user requirements are usually assumed to be homogenous and used mainly to define the conceptual model of a $\mathcal{D}\mathcal{R}$. In this paper, we study the interest of the requirements in the other phases of the life cycle of a $\mathcal{D}\mathcal{R}$. We propose a generic model based on ontologies to unify the used vocabularies and requirements languages. Then we extend this model using the formal method B to check the consistency of the requirements w.r.t. the integrity constraints defined on the logical schema. Finally we propose to select optimization structures of a $\mathcal{D}\mathcal{R}$ using the user requirements instead of SQL queries. Several experiments on the Star Schema Benchmark (SSB) confirm the interest of our proposition.