Integrating Heterogenous Overlapping Databases through Object-Oriented Transformations
Vanja Josifovski, Tore Risch · Very Large Data Bases · 1999
Integration of data from autonomous and heterogeneous data sources often requires means to mediate and reconcile overlaps and con icts between the integrated data. It is also desirable that the mediator system stores local data associated with the data from the sources. Achieving acceptable query response time for a mediator system has been a known research problem. This work presents a mediator query processing framework based on a representation of the data mediation and reconciliation by a number of auxiliary system-de ned object-oriented (OO) views and overloaded functions (queries). The framework is supported by de ning an overloading and late binding mechanism for the OO views through declarative queries. A query over the mediated OO views will have late bound subquery invocations which are transformed into disjunctive query expressions. Consistency and completeness of the queries are guaranteed by expanding the queries with validation subqueries. Performance is improved by type aware query rewrites and selective OID generation in the mediators. Experiments show that the proposed query optimization dramatically improves the query execution time compared to a naive instance-oriented query strategy or partial strategies.