Distributed data mining(DDM) is widely used in industrial,scientific and commercial applications to analyze large datasets maintained over geographically distributed sites,which makes DDM a major research issue on(today's) data mining field.This paper discusses the disadvantages of existing DDM systems,puts forward a service-oriented architecture for DDM on the grid and describe the process of mining under this framework.In the proposed,serviceoriented,cross-platform framework,the mining algorithm and distributed data sets are packed into the Web Service Resources(WS-Resource) respectively,which can cooperate to perform DDM as required dynamically.Finally,a grid based on local area network was built with Globus Toolkit 4.0 and take an example of association rule mining to illustrate how to work with the proposed architecture.