Currently most distributed peer-to-peer network systems send to other peers when users' query request, but their search result is a great deal of needless answer as gap between their search goal and their information need. It is because each peers own different interest domain knowledge. However, we purpose interest-based peer selection in order to rank best peers under the user context with content retrieval tasks. Generally speaking, providers may be best peers what user want find. Providers can satisfy users' information need, if we figure out providers what are user search goal. Based on ACM domain knowledge, we propose reference ontology in order to calculate the semantic similarity under each peer preferences, and rank best peers. Feedback data is updated ontological user profile by user activities and content retrieval tasks. Our approach can satisfy users' information need and reduce their search time-consuming.