P2P systems are increasingly used to discover and share various data between users. The performance of a P2P based information retrieval system is determined by the efficiency of locating information and the precision with which the retrived data corresponds to the submitted query. In this paper, we address the problem of similarity search by providing a combined peer organizing and data organizing scheme for approximate similarity search in a Hybrid Overlay P2P Network. Through extensive simulations, we show the efficiency of the similarity search in our approach using a density-based algorithm to cluster data and peers in a high dimensional feature space.