Perspectives on database theory

Database management systems address the need to store, retrieve, and manipulate large amounts of data in an organized fashion. The database held has grown tremendously in the last 25 years. It is reported that the database industry generated $7 billion in revenue in 1994 and is growing at a rate of 35% per year. Industrial and academic research have been instrumental to this growth. Theory has played an important role in defining the right abstractions and concepts, and providing a firm foundation for the field. In order to access effectively a large volume of data, one needs an abstract logical view of the data, which must be separate from the physical storage of data. The important first component of a database is therefore an abstract view of data (called the data model) and the accompanying specialized high-level language that is used to access the data. The second important component is the data structures that are used to store the data along with the algorithms to support the efficient translation from the logical to the physical world. The third important component is the mechanisms that allow the database to be accessed concurrently by many users, without violating its integrity. Theory has contributed to all three fronts, starting with what is undoubtedly the cornerstone of the area, the introduction and formal definition of the relational model by F.P. Codd (1970). It is a highly unusual compliment for theory when the major commercial products in the field have at their core a mathematically rigorous, formal model. Our primary aims in this paper will be to give a flavor of the types of problems that database theory addresses, and to review how research in the area has evolved over the years. At the end we will try to point to some topics that may be of interest to people in the FOCS community tempted to work in database theory.