As the applications become more and more complicated, the requirement of Web service composition will be more urgent. In this paper, we put forward a new storage strategy for web services which can be flexibly extended in relational database. We also present a matching algorithm SMA between web services of multiple input/output parameters, which considers the semantic similarity of concepts in parameters based on WordNet. Moreover, a service composition algorithm Fast-EP based on the above storage strategy is presented. Because utilizing the characteristic and using index mechanism in relational database, we obtain highly efficient web service composition. We extract the feature vector of each web service or composite service, and create dynamic linear hash index on these vectors so that the results of each search could be hierarchical classified. QoS and service price are utilized to rank the result set. At last, we develop a Web services search engine WSIS and show through experiment that our approach in this system has better efficiency of service composition and higher recall ratio of service search than traditional approaches.