With the development of the Web, an information "Big Bang” has taken place on the Internet. Search engines have become one of the most helpful tools for obtaining useful information from the Internet. However, instead of caring about the semantics of information, the machine on the current Web cares about the location and display of information only. Because of this shortcoming of the current Web, the search results by even the most popular search engines cannot produce satisfactory results. The development of the next generation Web, Semantic Web, will turn the situation around completely. This paper proposes a prototype relation-based search engine, "OntoLook,” which has been implemented in a virtual Semantic Web environment in our lab. We also present its system architecture and analyze the key algorithm.