Searching in a dataset remains a fundamental problem for many applications. The general purpose of many similarity measures is to focus the search on as few elements as possible to find the answer. The current indexing techniques divides the target dataset into subsets. However, in large amounts of data, the volume of these regions explodes, which will affect search algorithms. The research tends to degenerate into a complete analysis of the data set. In this paper, we proposed a new indexing technique called GHB-tree. The first idea, is to limit the volume of the space. The goal is to eliminate some objects without the need to compute their relative distances to a query object. Peer-to-peer networks (P2P) are superimposed networks that connect independent computers (also known as nodes or peers). GHB-tree has been optimized for secondary memory in peer-to-peer networks. We proposed a parallel search algorithm on a set of real machine. We also discussed the effectiveness of construction and search algorithms, as well as the quality of the index.