application repository

• name: China Meteorological Application Grid (CMAG)
• domain: Meteorology
• country: China
• author:
• institute: Chinese Academy of Meteorological Sciences
• contacts:
• description: China Meteorological Application Grid (CMAG) is built on the CMA dedicated communication system and distributed computing resources, covering regional centers in Beijing, Guangdong, Shanghai, Wuhan and Guangxi province. The aggregated computing power has reached Tera-Flops scale. It also achieves interconnection and sharing of computing resources and NWP software among the Chinese Academy of Meteorological Sciences, the National Meteorological Information Center, and other related regional centers. CMAG will eventually become an environment for operational NWP services and for cooperative NWP research and development.
  CMAG also provides new time- and-location-specific weather forecast services for areas where computing resources are in shortage, for example, Qinghai province and Yantai city.  
• functionalities:
  Via the web portal http://grid.cma.gov.cn, unauthorized clients can view the prediction results, while registered users can submit single or multi-member ensemble jobs to the aggregated computing resources, display the visualized products, modify or improve the forecast system in a collaborative environment. The development of grid ensemble prediction system has been completed and part of the results are provided on the web; however works are still ongoing and only a part of the computing resources have been aggregated and users can utilize the platform to submit numerical prediction jobs.
  Some of the functionalities are in process of being developed or foreseen to be developed, and may not be available yet. The following is a set of relevant features currently under development:
  • Grid Ensemble Prediction Application System. a grid ensemble prediction system based on GRAPES meso-scale model, and focusing on initial perturbation and model perturbation, suitable for climatic and terrain characteristics of mainland as well as the assessment of ensemble prediction system.
  • Data flow management focusing on management of heterogeneous and massive meteorological data on grid environment, including location, collection, transmission, storage and visualization of these data.
  • Ensemble prediction system Portal . Based on CMAG Portal, and concerning complex data flow and huge computation of the ensemble prediction system, an ensemble prediction system operation control interface is being developed to allow researchers to perform experiments at anytime and anywhere.
  • Collaboration research environment. Considering the fact that mesoscale ensemble prediction researchers are scattered around the country, a collaborative environment will be established to provide a grid testbed for numerical prediction research and development.
  • Grid computing platform. The geographically distributed computing resources will be aggregated, including those of Chinese Academy of Meteorological Sciences, Meteorological Information Center, Shanghai Typhoon Institute, Wuhan Institute of Heavy Rain, Chengdu Institute of Plateau Meteorology, Shenzhen Meteorological Bureau, Guangxi Meteorological Bureau as well as CNGrid, to provide a platform for ensemble prediction and research.
• middleware requirements:
  • Sharing of software. Atmospheric numerical modelling is the process of solving a set of equations to obtain an objective forecast of the future state of the atmosphere. The equations describe the evolution of many variables and together define the state of the atmosphere. Major software shared: meso-scale model GRAPES, ensemble prediction system GRAPES-EPS
  • Sharing of hardware. The computing resources in meteorological departments are distributed geographically and heterogeneously, such as IBM, SW, Dawning, etc., so the collaboration and sharing of these resources are a key problem.
  • Data storage and data sharing. Massive data is involved in meteorological forecast, which includes data assimilation, multi-member ensemble prediction perturbation, corresponding ensemble prediction products, visualization of the results, etc. The access, formats and storage locations of these data may be different, and this makes the system extremely complex in data management and utilization. Therefore, a unified data management system should be constructed.
  • High-performance computing and collaborative research. A collaborative environment is required to provide a shared computing platform to modify the source codes, to improve numerical schemes and to submit jobs via web browser.
• resources requirements:

Back to list of applications >>

powered by