Semantic Web services (SWs) has become the most dominant paradigm of the service-oriented computing and one of the hot issues in the area of distributed computing technology to perform business services composition more efficiently and effectively for a number of years now. The distributed composition of SWs according to their functionality increases the capability of an application to fulfill the user’s requirements. In this paper, we describe an efficient approach for improving the performance and effectiveness of automatic and cooperative composition of SWs in P2P systems. It implements a distributed solution based on scalable epidemic algorithm to discover and compose SWs in P2P systems. The main idea of our approach is to develop hybrid matching technique that operates on OWL-S process models in order to ensure high recall, further reduce the number of messages exchanged and reduce the execution time for discovering and composing SWs in the P2P network. Moreover, our matching technique is able to detect complex matching between these SWs based on their parameters and the user request. We propose a similarity measure that will be used to compose new discovered and heterogeneous collaborative Web services of large-scale distributed systems in a P2P network for satisfying user requirements, and to rank the results according to a similarity score expressing the affinities between each of them and a user-submitted query. The experimental results show that our approach is efficient and able to reduce considerably the execution time and message overhead, while preserving high levels of the distributed discovery and composition of SWs on large-size P2P networks.