基于MapReduce的CME参数识别模型并行计算技术

日冕物质抛射（Coronal Mass Ejection，CME）参数识别模型是太阳风预报过程的重要组成部分.在空间环境预报业务中，为提高太阳风预报的准确率，需要提高CME参数识别的精度.模型以计算任务串行的方式运行，运算效率低导致模型运算时间长，不能满足这种需求.CME参数识别模型的物理运算过程相互不独立，其在单节点上的运行方式不能满足并行化要求.基于MapReduce的并行计算框架，改进了CME参数识别模型的计算流程，提出CDMR（CME detection under MapReduce）方法，实现了CME参数识别模型的并行计算，并对比分析CME参数识别模型在串行计算和MapReduce并行计算下的运行时间，提高了模型的识别精度和计算效率. 

Parallel Computing Technology for CME Parameter Detection Model Based on MapReduce

Space environment prediction model is an important part of space environment business.Coronal Mass Ejection(CME)is the source of many space events and near-Earth space environment disturbances.The CME parameter detection model is an important part of the solar wind forecasting process.In order to improve the accuracy of solar wind forecasting in space environment forecasting,it is necessary to improve the accuracy of CME parameter detection.However,the model runs in serial mode with low calculating efficiency which leads to long operation time of the model and can not meet the requirement.Based on the parallel computing framework of MapReduce,according to the characteristics of CME parameter detection model,the calculation flow of CME parameter detection model is improved.A CDMR(CME Detection under MapReduce)method is presented,which can realize the parallel computing of CME parameter detection model.Moreover,the running time of the CME parameter detection model between serial computing and MapReduce parallel computing is compared.The experimental results show that the running time is reduced by using MapReduce parallel computing,and the detection accuracy and calculation efficiency of the model are improved.