Marine Environmental Numerical Prediction Data Processing Method Based On The Construction Of Synergistic And Parallel Computing

As safeguard of various marine activities, the marine environmental forecast is indispensably significant. Gradual improvement of marine environmental forecasting techniques has be spawning a large number of numerical prediction data. But in the current marine environmental forecasting work, there are some limitation in expansion of data storage, processing performance, forecasting information service mode and so on. Combining theoretical and technical analysis of cloud services and high-performance computing, the research work of this dissertation is aimed to build an integrated and efficient service architecture for marine environmental forecasting information, which is used to store and organize the heterogeneous forecasting data in a unified way, and supplys the efficient method of parallel computing and processing the data. Specific study as follows:(1)The extensible model of multi-layered service pool for parallel computing the marine environmental forecasting data. Based on the uniform interface specification, the dissertation designs the highly applicable service pool made up with multi-granularity general service units, which solve the problem of fixed and single service mode. The heterogeneous data and service resources can be effectively integrated and flexibly extended so that the multi-layered service pool is able to provide the fully open-ended service mode.(2) The unified method for storage and organization of marine environmental forecasting data based on HBase and HDFS. A model of the tile pyramid is put forward to store and organize the data in a unified way. Combining with the module, HBase and HDFS could store and manage the numerical prediction data after stratification and partitioning, which can avoid the limitation of relational database and make it easy to utilize multiple data synthetically.(3) The mutilevel construction model of synergistically and parallel computing drived by the service. Based on the unified method of data management, HDFS and HBase are merged with MapReduce parallel computing framework and MATLAB technical computing environment. The processes of data integration and parallel computing drived by the service are implemented based on the construction model, which overcomes the bottleneck of distributed data processing.(4) Constructing the integrated service platform for marine environmental forecasting information. Relying on the multi-layered service pool and mutilevel construction of synergistically and parallel computing, and integrating distributed storage and high-performance parallel computing in marine environmental forecasting data storage and management, a prototype platform is developed. Ultimately, the dissertation implements the visualization client application using Web services, RIA(Rich Internet Applications) technique and FSH (Flex+Spring+Hibernate) architecture, which realizes the extension according to the needs and quick response of the forecasting information service.