Selectively materializing data in mediators by analyzing user queries

There is currently great interest in building information mediators that can integrate information from multiple data sources such as databases or Web sources. The query response time in such mediators is typically quite high. We present an approach for optimizing the performance of information mediators by selectively materializing data. We first present our overall framework for materialization in a mediator environment. We then stress the need to materialize data selectively and outline the factors that must be considered to select data to materialize. We present an algorithm for identifying classes of data to materialize by analyzing one of the factors which is the distribution of user queries. We present results with an implemented version of our optimization system for the Ariadne information mediator which show significant performance improvement over no materialization and existing schemes such as page level caching