Relational Databases

Abstract

Database management systems (DBMSs) aim to provide users with an efficient tool for good modeling and for easy and efficient manipulation of data. The first DBMSs were based on a hierarchical and then network-like conceptual data organization, which actually reflects the physical organization of the underlying files. These systems, although efficient, have some important drawbacks. The relational model, proposed by Codd in 1970, avoids these drawbacks by distinguishing explicitly between the physical and the conceptual levels of data organization. This basic property of the relational model is a consequence of the fact that, in this model, users see the data as tables. The tables of a relational database are accessed and manipulated as a whole, contrary to languages based on hierarchical or network models, according to which data are manipulated on a record-by-record basis. As a consequence, data manipulation languages for relational databases are set-oriented, and so, they fall into the category of declarative languages. On the other hand, because relations are a well-known mathematical concept, the relational model stimulated a lot of theoretical research that led to successful implementations. In this article, we focus on the theory of the relational model and on basic aspects of dependency theory. Then, we address problems related to updates and transaction management, and we briefly describe the structure of relational systems and the associated reference language called SQL. We conclude with a brief discussion on extensions of the relational model.