Under the joint influence of high-intensity human activities and cli-mate change,the coastal ecological environment is deteriorating,and the ecologi-cal environment security and the sustainable development of the marin...Under the joint influence of high-intensity human activities and cli-mate change,the coastal ecological environment is deteriorating,and the ecologi-cal environment security and the sustainable development of the marine economy are seriously threatened.Therefore,it is of great significance to establish a high-resolution ecological environment operational forecasting system.To meet the run time requirements of the ecological operational forecasting system,a vari-ety of parallel optimization methods were proposed to improve the operation efficiency of the model.First,based on the National Marine Environmental Fore-casting Center’s Lenovo cluster,the ROMS benchmark experiment was expanded to the 4000 Processes scale.A good speedup was obtained by the experiment.The ROMS model was analysed with strong scalability.Second,in the hydrodynamic-ecological simulation experiment of the Bohai Sea-Yellow Sea-East China Sea,by optimizing Vector,InfiniBand,and Parallel I/O,the performance of the model can be improved by 270%while maintaining the same computing resources.That computing resources were more reasonably used lay the foundation for the operational forecast.展开更多
基金This research is supported by the National Natural Science Foundation of China(41976200)the project of Guangdong Ocean University(060302032106)Southern Marine Science and Engineering Guangdong Laboratory(Zhuhai)(SML2022SP301)。
文摘Under the joint influence of high-intensity human activities and cli-mate change,the coastal ecological environment is deteriorating,and the ecologi-cal environment security and the sustainable development of the marine economy are seriously threatened.Therefore,it is of great significance to establish a high-resolution ecological environment operational forecasting system.To meet the run time requirements of the ecological operational forecasting system,a vari-ety of parallel optimization methods were proposed to improve the operation efficiency of the model.First,based on the National Marine Environmental Fore-casting Center’s Lenovo cluster,the ROMS benchmark experiment was expanded to the 4000 Processes scale.A good speedup was obtained by the experiment.The ROMS model was analysed with strong scalability.Second,in the hydrodynamic-ecological simulation experiment of the Bohai Sea-Yellow Sea-East China Sea,by optimizing Vector,InfiniBand,and Parallel I/O,the performance of the model can be improved by 270%while maintaining the same computing resources.That computing resources were more reasonably used lay the foundation for the operational forecast.
