The most well-known use of service-oriented architectures has given rise to several challenging research issues. One of these is the ability to recursively create a composite service as a process of other already available services, which are created by various companies and offer a variety of functionalities. Due to the ever -growing number of functionally identical services supplied on the web, it is crucial to identify them using well defined non -functional criteria. This needs an efficient method of selecting such services to compose. In this paper, adopting a service process and directed acyclic graph, we present an efficient and scalable approach to address the issue. Services are chosen for their composition based on both positive and negative QoS qual ities, aiming to optimize user satisfaction. Experimental results show that our proposal scales very well regarding the huge amount of candidate services. It is less complex, and it depends linearly on the services implied in the composition.