Object-oriented modeling and design of database federations

We describe a logical architecture and a general semantic framework for precise specification of so-called database federations. A database federation provides for tight coupling of a collection of heterogeneous component databases into a global integrated system. Our approach to database federation integrates in a uniform and systematic manner the underlying database schemas of the component legacy systems to a separate, newly defined integrated database schema. This integrated database is completely virtual, and will constitute the actual federated database. That is, queries posed against the federated system will be posed against this virtual integrated database; these global queries will then be mapped by the mediator to actual local queries against the existing (legacy) component databases. Our approach is based upon the UML/OCL data model. UML is the de facto standard language for analysis and design in object-oriented frameworks, and is being employed more and more for analysis and design of Information systems, in particular information systems based on databases and their applications. Database specifications often involve specifications of constraints, and the Object Constraint Language (OCL) - as part of UML - can aid in the unambiguous modelling of database constraints. One of the central notions in database modelling and in constraint specifications is the notion of a database view; a database view closely corresponds to the notion of derived class in UML. We will employ OCL and the notion of derived class as a means to treat (inter-)database constraints and database views in a federated context. We will also offer a transaction model for a simple set of updates in database federations. The paper will demonstrate that our particular mediating system integrates component schemas without loss of constraint information. Furthermore, we will discuss a mapping of database specifications in terms of UML/OCL to the relational model