Validating Mediator Cost Models with Disco

Disco is a mediator system developed at INRIA for accessing heterogeneous data sources over the Internet. In Disco, mediators accept queries from users, process them with respect to wrappers, and return answers. Wrapper provide access to underlying sources. To efficiently process queries, the mediator performs cost-based query optimization. In a heterogeneous distributed database, cost-estimate based query optimization is difficult to achieve because the underlying data sources do not export cost information. Disco's approach relies on combining a generic cost model with specific cost information exported by wrappers. In this paper, we propose a validation of Disco's cost model based on experimentation with real Web data sources. This validation shows the efficiency of our generic cost model as well as the efficiency of more specialized cost functions.