The paper identifies various problems in integrating schema, data and methods in heterogeneous systems and describes the strategies used in Pegasus to deal with them. Pegasus is an object-oriented multidatabase system that is being developed at Hewlett-Packard Laboratories. The goal of the first running prototype is to provide facilities for multidatabase applications to access and manipulate multiple autonomous heterogeneous object-oriented and relational databases. Pegasus provides both type and procedural abstractions to support integration. Type abstraction is used to organize objects from multiple databases in categories that are related by having some common properties. The common properties of objects in different databases may be in incompatible forms or in inconsistent states. The system provides procedural abstraction that enables the schema integrator to develop arbitrary mappings between data from different sources and to write necessary transformations to convert object properties to a common form. Because of the importance of existing relational databases, some important integration techniques are demonstrated through the object-oriented views over SQL databases.
展开▼