面向时空大数据的高性能并行计算策略对比  

A comparation of high-performance parallel computing strategies for spatiotemporal big data

作　　者：王思敏 王斌[1,2,3] 徐佳沅 陈旭[1,2,3] WANG Simin;WANG Bin;XU Jiayuan;CHEN Xu(Surveying and Mapping Institute of Land and Resources of Guangdong Province,Guangzhou,Guangdong 510663,China;Key Laboratory of Natural Resources Monitoring in Tropical and Subtropical Area of South China,Ministry of Natural Resources,Guangzhou,Guangdong 510663,China;Guangdong Geographical Science Data Center,Guangzhou,Guangdong 510663,China)

机构地区：[1]广东省国土资源测绘院,广东广州510663 [2]自然资源部华南热带亚热带自然资源监测重点实验室,广东广州510663 [3]广东省地理科学数据中心,广东广州510663

出　　处：《北京测绘》2025年第4期456-461,共6页Beijing Surveying and Mapping

基　　金：自然资源部华南热带亚热带地区自然资源监测重点实验室开放基金(2022NRM006);广东省省级科技计划(2021B1212100003)。

摘　　要：测绘地理信息数据作为国家重要资源,其数据量级在不断增长,空间矢量数据的复杂结构促使高性能并行计算成为必然需求。当前研究多侧重单一数据模型和固定计算模式,综合分析对比等研究较少。本文提出高性能空间分析并行计算策略,将其分解为数据、算子、并行、环境四类综合模块并设计相应策略。通过试验对比,结果表明在计算密集型和数据密集型场景中,并行计算优于传统单线程计算。分布式并行计算架构性能提升显著,空间数据库并行计算策略在处理大型数据时优势明显,能支持并发操作解决关键问题。同时,空间计算分析性能与中央处理器(CPU)核数并非完全线性相关,受磁盘输入输出(I/O)资源限制和任务数据依赖性影响。本文为测绘地理信息服务提供了高性能计算和数据处理的理论依据和实践参考。Geospatial information data,as an important national resource,is continuously growing in volume.The complex structure of spatial vector data necessitates the demand for high-performance parallel computing.Current research often focuses on single data models and fixed computing modes,with fewer studies on comprehensive analysis and comparison.This paper proposed a high-performance spatial analysis parallel computing strategy,which was broken down into four integrated modules:data,operators,parallelism,and environment,with corresponding strategies designed for each.Experimental comparisons show that in both compute-intensive and data-intensive scenarios,parallel computing outperforms traditional single-threaded computing.The performance of distributed parallel computing architecture is significantly improved,and the parallel computing strategy for spatial databases demonstrates clear advantages in handling large-scale data,and supporting concurrent operations to address critical issues.Moreover,the performance of spatial computing and analysis is not strictly linearly related to the number of central processing unit(CPU) cores,as it is influenced by disk input/output(I/O) resource limitations and task data dependencies.This paper provides theoretical foundations and practical references for high-performance computing and data processing in geospatial information services.

关 键 词：高性能并行计算 时空大数据 分布式计算架构 对象关系型数据库 

分 类 号：P208[天文地球—地图制图学与地理信息工程]