For the problem of semantic content sharing of information resources in P2P networks, this paper first proposes a multiple-ontology based resource metadata description framework, which divides resource metadata into core metadata and extensional metadata. Then, based on the metadata framework, this paper places emphasis upon the research of indexing strategy of resource metadata. A hybrid indexing architecture is provided. It uses community strategy to organize core metadata element and uses DHT and semantic association to index extensional metadata element, which combines local indexing and global indexing. Based on this indexing strategy, the work flow of resource discovery and the load balance strategy are designed. Compared with the related current systems, this indexing architecture will greatly improve the efficiency of resource discovery and the system's scalability.