This paper presents the process for research and implementation a real-time duplicate software in DB2 database. This software used the modular design, and has 5 major modules: Client, Server, Full synchronizing, Increase synchronizing, Load. While the software running, the 5 modules coherent on with another. The client call the server, and the server schedules all task. In the process of the implementation, we grap the database's log real-time to gain the change. This software doesn't use any DB2 database's backup interface, so it well done in the function and efficiency. The test environment is OS:LINUX, CPU:Core 2.4 GHz ,Memory: 2 GB. Under such a test environ-ment, this software worked well.