20.3. Integrated or federated systems and multi-databases

The direction of integrated or federated systems of heterogeneous databases and multi-databases appeared due to the need to integrate database systems based on different data models and managed by different DBMS.

The main task of integrating heterogeneous databases is to provide the users of the integrated system with a global database schema, represented in some data model, and automatically convert global-level database manipulation operators into operators understandable by the corresponding local DBMS. In theoretical terms, transformation problems have been solved, there are implementations.

With strict integration of heterogeneous databases, local database systems lose their autonomy. After the inclusion of the local database in the federated system, all further actions with it, including administration, should be conducted at the global level. Since users often do not agree to lose local autonomy, wishing nevertheless to have the opportunity to work with all local DBMS in the same language and formulate queries with simultaneous indication of different local DBs, the direction of multi-DB is developing. In multi-database systems, the global integrated database schema is not supported and special naming methods are used to access local database objects. As a rule, in such systems at the global level, only data sampling is allowed. This allows you to maintain the autonomy of local databases.

As a rule, it is necessary to integrate heterogeneous databases distributed in a computer network. This greatly complicates implementation. In addition to its own integration problems, it is necessary to solve all the problems inherent in distributed DBMS: global transaction management, network query optimization, etc. It is very difficult to achieve efficiency.

As a rule, for the external representation of integrated and multi-DB, a (sometimes extended) relational data model is used. Recently, it has increasingly been proposed to use object-oriented models, but in practice, while the basis is the relational model. Therefore, in particular, the inclusion of a local relational DBMS in an integrated system is much simpler and more efficient than the inclusion of a DBMS based on a different data model.