Heterogeneous databases and high level abstraction

A heterogeneous database management system combines multiple dissimilar models of data within a single integrated system. The objective is to allow a user to access data independently of how it is actually organized. For example, a user may access a database as though it were stored relationally (i.e., in tables) [CODD70], even though it is actually stored as a CODASYL/DBTG or network database [CODA71]. In addition, different subpieces of the database may be organized under different data models. The heterogeneous database system must present these to the user as an integrated whole. The user's model of his data may be different from any of the models chosen to implement it. Rather than construct a new database system from scratch, we are interested in constructing a heterogeneous system out of existing systems. The key difficulties with this approach are: (1) the formulation of database design methods that are applicable to a variety of different data models, and (2) the development of techniques to translate programs and data between dissimilar data models. In this paper, we briefly describe how high level abstraction has been applied to these problems. The use of abstraction in database systems is related to the application of abstraction techniques in programming languages and artificial intelligence research.