天气雷达组网拼图并行处理的目的就是为了满足实时业务的需求,通常的串性方法在雷达运行的六分钟周期内不能完成。利用共享存储多核处理器的高性能服务器,研发了高频度实时全国天气雷达资料处理的并行计算系统,根据单部天气雷达处理和多部雷达拼图的特点,提出了进程级别和线程级别细粒度的OpenMP共享内存编程的两级并行方法,给出一个合理的派生线程粒度数,同时利用共享内存文件的方法提高程序的执行效率。实验结果表明,使用两级的并行计算等方法可以满足全国天气雷达实时处理的要求。利用并行计算实现实时性要求高的业务系统是重要的手段。
Weather radar mosaic parallel processing purpose is to meet the demand of real-time service, usually serial method cannot achieve in radar running cycle. Using high performance server of shared memory multiprocessor, develop the parallel computing system of high frequency real-time weather radar data processing. According to single weather radar processing and multi-radar mosaic characteris- tics, put forward the two stage parallel method of the process level and thread level of fine particle size OpenMP shared memory program- ming. Give a reasonable derived thread size, put forward the shared memory file method to improve the program execution efficiency. The experimental results show that, the use of two stage parallel calculation method to the national weather radar can satisfy the requirement of real-time processing. Use of parallel process is important for real-time system.