Optimizing Queries for Repository-Based Applications
Query optimization strategies for repository systems must take into account the rich and often unpredictable structure of metadata, as well as supporting complex analysis of relationships between those structures. This paper describes rationale and design of a cost-based query optimizer offered in ConceptBase, a metadata manager that supports these capabilities by a deductive object-oriented data model. In contrast to most implemented DBMS, the optimizer is not based on the concept of join selectivity, but on detailed distribution information about object and literal fan-out, with heavy exploitation of materialized views and end-biased histograms. Experiences from two real-world applications in software configuration management and cooperative design demonstrate the practical advantages but also point to some lessons for further improvement.
ACM Symposium on Applied Computing, San Antonio, Texas, USA, 1999