In this paper, we study the problem of indexing multi-dimensional data in the P2P networks based on distributed hash tables (DHTs). We identify several design issues and propose a novel over-DHT indexing scheme called m-LIGHT. To preserve data locality, m-LIGHT employs a clever naming mechanism that gracefully maps the index tree into the underlying DHT so that it achieves efficient index maintenance and query processing. Moreover, m-LIGHT leverages a new data-aware index splitting strategy to achieve optimal load balance among peer nodes. We conduct an extensive performance evaluation for m-LIGHT. Compared to the state-of-the-art indexing schemes, m-LIGHT substantially saves the index maintenance overhead, achieves a more balanced load distribution, and improves the range query performance in both bandwidth consumption and response latency.