面向海量数据的相对波速变化计算的并行化方法  

A PARALLELIZATION METHOD OF RELATIVE WAVE VELOCITY VARIATION CALCULATION FOR MASS DATA

作　　者：张赛 司冠南[1] 周风余[2] 蔡寅[3] Zhang Sai;Si Guannan;Zhou Fengyu;Cai Yin(School of Information Science and Electrical Engineering,Shandong Jiaotong University,Ji’nan 250300,Shandong,China;School of Control Science and Engineering,Shandong University,Ji’nan 250100,Shandong,China;Shandong Earthquake Agency,Ji’nan 250014,Shandong,China)

机构地区：[1]山东交通学院信息科学与电气工程学院,山东济南250300 [2]山东大学控制科学与工程学院,山东济南250100 [3]山东省地震局,山东济南250014

出　　处：《计算机应用与软件》2022年第2期21-25,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(61375084);国家重点研发计划项目(2017YFB1302400);山东省自然科学基金项目(ZR2019MF064);山东省重点研发计划项目(2017GGX10119)。

摘　　要：随着地震台站数量大大增加,测量数据量也急剧增长。传统的串行化相对波速变化计算方法面向海量数据时存在计算速度慢、消耗时间长等问题,已不能满足日常业务的需求。针对此问题,提出一种面向海量数据的相对波速变化计算的并行化方法。通过对地震数据集的划分和算法调度,将数据集分布到基于Spark计算框架的分布式集群上进行并行运算。实验表明,该方法提高了相对波速变化计算的速度,显示出计算模型的稳定性和可扩展性。With the large increase in the number of seismic stations,the amount of measured data has also increased dramatically.Traditional serialized relative wave velocity calculation methods have problems such as slow calculation speed and long time consumption when facing massive data,and they cannot meet the needs of daily business.Aiming at this problem,a parallelization method of relative wave velocity variation calculation of massive data is proposed.The data set was distributed to a distributed cluster based on the Spark computing framework for parallel operation by dividing the seismic data set and algorithm scheduling.Experiments show that this method improves the calculation speed of the relative wave velocity variation,and the calculation model has stability and scalability.

关 键 词：相对波速变化 并行计算 海量数据 分布式 归一化 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]