Extending Relational Query Optimization to Dynamic Schemas for Information Integration in Multidatabases

Abstract

This paper describes an implementation of FISQL/FIRA for multidatabases. FISQL/FIRA are transformationally complete relational query languages that fully support data-metadata transformations within the relational model. The system can facilitate many known data integration architectures, such as GAV, LAV, or GLAV. In addition to fully supporting relational optimization heuristics, our implementation uses a bi-level optimization paradigm where purely relational sub-fragments of queries may be pushed into source query engines. This paradigm shares features of canonical distributed database processing, but our framework has a new dimension based on the potential of unwinding query metavariables based on global metadata information. We discuss tradeoffs of the approaches and present initial experimental results for the system.