Currently most centralized RDF storage and query architectures have problems in query performance and scalability. In this paper we present a distributed RDF query architecture to address these problems, which is built on Chord and Sesame. In our architecture, we construct semantic topology with distributed interestindex to route query and take advantage of semantic similarity between query and peer to select peers which are capable of answering the given query. Experimental results have shown that our approach has much better performance while achieving a good recall rate.