Abstract
Nowadays, we see an explosion in the number of Database Management Systems (DBMS) in the market. Each one has its own characteristics. This spectacular development of DBMS is mainly motivated by the need for storing and exploiting the deluge of heterogeneous data for analytical purposes. As a consequence, companies and users are faced with huge range of choices and sometimes it is hard for them to find the relevant DBMS. Some Web sites such as DB-Engines (http://db-engines.com/en/) provide monthly a classification of hundreds of DBMS (303 in April 2016) using metrics related to usage and user feedbacks. These criteria are not always sufficient to help companies and users to make a good choice. Therefore, they have to be enhanced by qualitative measurements obtained by testing the activities of DBMS for a set of non-functional requirements. In this perspective, some council such as Transaction Processing Council publish non-functional requirement results of DBMS using their own benchmarks. Another serious producer of test data is the researchers via their scientific papers. Each year they publish a large amount of results of new solutions. To facilitate the exploitation of these test results by small companies and researchers from developing countries, the construction of a test data repository connected to recommender system is an asset for companies/users. In this paper, we first propose a repository for structuring and storing test data. Secondly, a recommender system is built on the top of this repository to advise companies to choose appropriate DBMS based on their requirements. Finally, a proof of concept of our recommender system is given to illustrate our proposal.