Research On Parallel Computing Technology Of CME Parameters Detection Model Based On MapReduce

With the continuous development of science and technology,scientific computing,industrial applications,global climate forecasting,space environment business,human genetic engineering and other important research topics challenge the traditional serial computing,and put forward higher demand for computing efficiency.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.When a CME reaches the earth,it may cause strong geomagnetic storms,which will affect the safe operation of spacecraft,orbit maintenance,ground power system and so on to varying degrees.Accurate prediction of solar n is an important factor in geomagnetic storm prediction.Rapid and accurate identification of CME parameters is a very important factor in predicting whether and when CME can reach the Earth and a prerequisite for effective prediction of solar wind.At present,in the case of serial computation of computing tasks,the low efficiency of parameter identification model results in long operation time and low accuracy of parameter identification,which can not meet the needs of accurate prediction of solar wind.The single-node and multi-process concurrent mode can improve the efficiency of computing tasks to a certain extent,but it is limited by the limitation of single-node computing resources.Traditional parallel computing mode of multi-node data can solve the problem of limited computing resources,but due to the lack of distributed file system,there will be a large IO problem when sharing data sources,which affects the efficiency of reading and writing.MapReduce is a programming model suitable for distributed parallel computing,and distributed file system can provide distributed storage of data,which can solve IO problem well.Therefore,this paper proposes a model parallel computing method based on MapReduce to improve the efficiency of the model.The main work of this paper is as follows:(1)The basic concepts and development of parallel computing are investigated,and several typical parallel computing methods are compared and analyzed,including key technologies,computing processes,advantages and disadvantages,and applicable scenarios.According to the characteristics of spatial environment data and model,MapReduce is selected to carry out parallel design.(2)Aiming at the parameter identification model,because the physical operation process of the model is not independent of each other,and its running mode on a single node does not meet the requirements of parallelization,the application requirements of MapReduce technology oriented to the model's CPDM computing process were parallelized,and a parallel computing method based on MapReduce was proposed.This method greatly improves the computational efficiency of the model.(3)The parallel computing method based on MapReduce is extended to other space environment models,and a multi-space environment prediction model operation system based on MapReduce is designed.The system can realize the parallel operation and task scheduling of the model,and improve the computational efficiency of the model.