In the processes of ocean numerical simulation, with the increase in the calculation domain as well as the requirement of higher accuracy, a dramatic amount of data output make the I/O efficiency become a bottleneck of the whole system performance. To solve this problem, the parallel I/O technology was used in this study to optimize the output module, and the comparison with the traditional serial I/O methods under various numbers of computing nodes and various computing scales was conducted. Experimental results demonstrate that the proposed method significantly reduces the execution time of output module.