Dynamic source selection in large scale mediation systems

This paper proposes ORS, an original strategy to reduce the number of data, sources to access during query evaluation in large scale mediation systems. ORS proceeds first selecting sources using extensional (data) information to discard useless sources and then validates the intentional (schema) information that each one is able to provide. The first step is based on location queries on some "well chosen" data sources, that previously had made a consolidation integration effort. This paper proposes ORS to improve querying semantic virtual objects whose instances are distributed across numerous data sources. Cost analysis and implementation in a grid context; are also presented.