One of the essential problems in P2P is the strategy for resource discovery. Related methods in unstructured P2P either depend on the flooding and its variations or utilize various indices, which results in too much overhead to forward messages or too expensive cost to maintain the indices. An adaptive, bandwidth-efficient and easily maintained search algorithm for unstructured P2P systems, PeerRank, is presented. The scheme utilizes the feedback from previous searches to probabilistically guide future ones. In addition, an effective caching and indexing mechanism is introduced, which remarkably enforces the search performance. The final simulation experiment shows that the strategy can remarkably improve the search efficiency with the small average path length, high success rates, very low bandwidth consumption, and the eminent adaptability to the change of hot resources.