Dynamic meta-data support for information integration and sharing across heterogeneous databases

In a multidatabase environment, meta-data plays a more important role than in a single database environment. The federated database system manipulates data from several data sources. Users hardly have enough knowledge about the underlying meaning of the global schema. Meta-data describes the data semantic information that can help users to interpret the data. It provides necessary information for both query processing and data integration in a federated database system. However, traditional data modeling techniques such as relational data model and semantic data model are not designed to support efficiently the definition, storage and manipulation of complex meta-data.; The dissertation investigates on how the schema of the data model can be extended to support the meta-data, how this meta-data can enhance the semantics of a database schema, support data integration and mediation services, and facilitate query processing in a federated database system. The dissertation develops the Mediation Data Model. The Mediation Data Model provides an extensible schema that supports the dynamic meta-data, and a framework for specifying mediation services within the data model. The dynamic meta-data describes different types of data semantic information, in addition to the traditional static meta-data. The dynamic meta-data allows each instance object of a class to have associated with it distinct meta-data, whereas static meta-data assigns the same meta-data to an entire class rather than instances of a class. The Mediation Data Model provides the association of multiple unit values to an instance of a data type. The concept of Multiple Unit Value (MUV) is implemented by means of the Unit-type Dynamic Meta-Data, the Unit Value Mediator and the extended DML. The MUV concept allows object attributes to have values expressed in convertible units.; The dissertation also develops the federated database system architecture and a working prototype, called InfoFED. InfoFED makes use of the Mediation Data Model as a common data model to provide the self-describing data as well as an intelligent data manipulation language to users, and to improve data integration by reducing the schema integration conflicts. InfoFED also provides a browser for both data and meta-data, an object-oriented federated schema, and the data source registration and management.