GM卫星测高数据海面时变校正程序的I/O并行优化研究  

I/O parallel optimization of sea surface time-varying correction program for GM satellite altimetry data

作　　者：傅游 徐方正 梁建国 FU You;XU Fangzheng;LIANG Jianguo(College of Computer Science and Engineering,Shandong University of Science and Technology,Qingdao 266590,China)

机构地区：[1]山东科技大学计算机科学与工程学院,山东青岛266590

出　　处：《山东科技大学学报（自然科学版）》2023年第4期92-100,共9页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：山东省自然科学基金项目(ZR2022MF274)。

摘　　要：海面时变校正是建立平均海平面模型数据处理过程中最关键的一步,可以消除或削弱海面时变的影响。本研究实现了时空客观分析法对GM卫星测高数据进行海面时变校正的串行程序,针对该程序运行耗时超长问题,研究其I/O特征,提出两种多进程I/O并行方案,并使用消息传递接口(MPI)文件视口函数对其进行优化。在高性能集群系统上的实验结果表明,合并再分配方案具有更好的负载均衡度;I/O并行优化后的合并再分配方案加速效果明显,最高开启448进程时,大数据量实验耗时9283.84 s,与56进程时相比,加速比为7.21,并且具有良好的强、弱可扩展性。Sea surface time-varying correction is the most important step in the data processing process of establishing mean sea level model,which can eliminate or weaken the sea surface time-varying effect.This study designed and implemented a serial program for the objective analysis of time and space to perform sea surface time-varying correction on GM satellite altimetry data.To address the problem of the program’s excessively long processing time,the I/O characteristics were investigated and two multi-process I/O parallel schemes are proposed,which were optimized by using the massage passing interface(MPI)file view function.Experimental results on a high-performance cluster system show that the merging and redistributing scheme has better load balance.When 448 processes are used,the large-scale data experiment takes 9283.84 s and the speedup ratio is 7.21 compared to that when 56 processes are used,indicating that the optimized merging and redistributing scheme has a significant acceleration effect as well as good strong and weak scalability.

关 键 词：卫星测高 海面时变校正 客观分析法 消息传递接口 并行计算 I/O性能 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]