In recent years, a new type of applications called “Mashup” has developed on the Internet. These applications can help end-users to collect and operate the distributed data sources on the Internet. However, the research on performance optimization of Mashup operation is lacking. To address this shortfall, this paper proposes a new approach, POMO, for performance optimization of Mashup operation. POMO engages progress in three areas. Firstly, it implements dynamic cache point selection through the tradeoff between operation cost and caching benefit. Secondly, it implements cache point reusing by B+ tree index. Thirdly, it implements cache point update through incremental transmission. Experimental results show that POMO reduces the cost of Mashup operation and improves the performance of Mashup operation in dynamic environment.