With the explosion of services in grid environment, it’s necessary to develop a mechanism which has the ability of discovering suitable grid services efficiently. This paper attempts to establish a layered resource management model based on the locality principle which classifies services into different domains and virtual organizations (VOs) according to their shared purposes. We propose an ontology-based search method applying the ontology theory for characterizing semantic information. In addition, we extend the UDDI in querying, storing, and so on. Simulation experiments have shown that our mechanism achieves higher performance in precision, recall and query response time. With the explosion of services in grid environment, it’s necessary to develop a mechanism which has the ability of discovering suitable grid services efficiently. This paper attempts to establish a layered resource management model based on the locality principle which classifies services into different domains and virtual organizations We recommend an ontology-based search method applying the ontology theory for characterizing semantic information. In addition, we extend the UDDI in querying, storing, and so on. Simulation experiments have shown that our mechanism achieves higher performance in precision, recall and query response time.