The synchronous communication technology used in Semantic Web services requires both the service owner and the invoker be available at the same time. Thus, their interaction reduces scalability. Triple Space Computing is envisioned as communication and coordination paradigm for Semantic Web Services, which is an extension of tuple space computing to support RDF. We use it for communication based on principle of persistent publication and read or write data. In this paper we describe an architecture based on Triple Space Computing for Semantic Web Services communication. The mechanism is based on asynchronous communication, which would greatly improve the performance and efficiency of communication.