It is an important problem to efficiently support similarity search for multi-dimensional data spaces in peer-to-peer (P2P) data management environment. Current unstructured P2P data sharing systems provide only a very rudimentary facility in query processing, i.e., matching-based query processing. This paper therefore presents a simple, yet effective index structure called EVARI (extended vector approximation routing index) to address the problem of multi-dimensional range search in unstructured P2P systems, by means of both data approximation and routing index techniques. With the aid of the EVARI, each peer can not only process range queries with its local dataset, but also route queries to promising peers with the desired data objects. In the proposed scheme, each peer summarizes its local content using space-partitioning technique, and exchanges the summarized information with neighboring peers to construct the EVARI. Furthermore, each peer can reconfigure its neighboring peers to keep the relevant peers nearby so as to optimize system resource configuration and improve system performance. Extensive experiments show the good performance of the proposed approach.