Today, data is increasingly important in the software of a wide variety of companies, so that the requirements in terms of volume, performance or storage have changed. As a result, the traditional relational model and the long-established database engineering processes are no longer sufficient. Indeed, other models, called NoSQL, have been developed to meet these new needs. These models, far from totally replacing the existing one, are on the contrary destined to co-exist in software ecosystems. These systems with multiple databases are called polystores. Because of this co-existence of models, the tasks considered complex in database engineering will be all the more complex. First, modelling, NoSQL systems allow a great variety of data representation relying on several data models. No current modelling language can unify these models while preserving the flexibility of representation of specific models. Secondly, the manipulation, themultiplication of databases implies, to query them, to know their own query language. Moreover, this requires the writing of complex join code in case of overlapping or duplication on distinct heterogeneous databases. And finally, the evolution, more models and more languages of manipulations are as many additional elements to be evolved in order to keep a functional system. In this thesis we try to bring solutions to facilitate the management of these three challenges by proposing a new unified modelling language as well as a conceptual data access code generator facilitating the manipulation and the evolution of hybrid polystores.
|la date de réponse||3 mars 2023|
|Superviseur||Anthony Cleve (Promoteur), Jean-Noel COLIN (Copromoteur), Vincent Englebert (Copromoteur), Wim Vanhoof (Président), Davide Di Ruscio (Jury) & Csaba Nagy (Jury)|