This work covers architectural issues and performance aspects of clusters of databases for usage in a decision support scenario. It presents a scalable, middleware-based cluster architecture and develops innovative algorithms for high-performance query routing based on approximated cache states.