Abstract: To quickly locate an appropriate web service from the large count of services on the Internet, a cluster method is adopted to manage services. The services with similar functions are defined as a service cluster and then the architecture oriented service cluster and service binding algorithm are presented. However, the granularity of service clusters impacts greatly on the efficiency of service discovery and aggregation. A three-dimensional granularity division method for service clusters is proposed in this study. It includes structure granularity, quantity granularity and quality granularity. And each kind of the granularity are divided into three levels, the division rules for each level are also presented. Simulation experiments are provided to illustrate the rationality of the proposed granularity division method.