Grid technologies and infrastructures facilitate distributed resource sharing and coordination in dynamic, heterogeneous, multi-institutional environments. A replica catalog is a Grid component that keeps replica locations of data objects and provides location transparency to data access. Replica selection is of great importance to data-intensive scientific computing targeted by many data Grid projects and the enabling virtual data technologies. We present here GRESS, a Grid replica selection service that provides a consistent interface to multiple replica selection implementations and supports flexible result delivery and easy configuration and deployment of new replica selection algorithms. GRESS is based on the Open Grid Services Architecture, which facilitates discovery and incorporation of the service by other Grid components such as Grid planners and virtual data workflow execution environments. We discuss the motivation for and implementation of GRESS and describe its role in a...