Indexing Attribute Domains in Relational DBMS

Abstract

Modern applications demand more flexibility in terms of data management, with which the rigid schemas of Relational Database Management Systems (RDBMS) have not properly coped. This demand has led to the development of DBMS that are not based on the Relational Model and, usually, do not employ SQL, the typical access language of the RDBMS. However, such DBMS, called NoSQL (Not Only SQL), stand only by the most prominent tools of their classes in the moment, and not by a conceptual foundation model. This leads to uncertainty in institutions that depend on such tools, because the fact that a tool is currently adequate does not guarantee that it will remain adequate or even exist in the future. Therefore, this project aims at developing data management solutions that meet the flexibility requirements of modern applications, yet maintaining the Relational Model as their conceptual reference. We advance from the premise that the rigidity of the current RDBMS derives from how they have been implemented and that it is possible to build an RDBMS that allows: (i) flexibility when defining a relation, so that the structure of the tuples need to be neither identical for the entire relation nor thoroughly defined in a tabular structure; (ii) flexibility when defining an index, so that it can be associated with attributes sharing the same domain across different relations, including attributes not predicted in the schema. In this context, this project proposes to define and develop the concept of domain index, in order to provide such flexibility for the current RDBMS. This project also involves to develop a prototype including all the modules needed to extend an existing RDBMS in order to execute and evaluate the new concept.