Peer-to-peer (P2P) computing is recognized as one of most prominent paradigms to achieve scalability in key components of Grid systems. One of these components is resource discovery, whose duty is to provide system-wide up-to-date information about resources, a task inherently limited in scalability. Unlike typical P2P systems, Grid systems manage not only static resources, but also resources whose characteristics change dynamically over time. To cope with this scenario, recent P2P-based Grid resource discovery systems employ a combination of Distributed Hash Tables (DHTs) to manage static resources, and unstructured (i.e., broadcast-like) search techniques for dynamic resources. In this paper we elaborate on this approach by designing and implementing a Grid resource discovery system that employs a dynamic querying algorithm over a structured DHT-based overlay. The system has been fully implemented and deployed on the Grid'5000 platform for testing and evaluation. The experimental performance results demonstrate the efficiency of the implemented system, both in terms of number of messages and time needed to complete the search.