分布式并行计算下GML空间数据复制同步更新机制研究

发布时间：2018-01-19 09:48

本文关键词： GML同步复制 GML更新 MongoDB GML ArcGIS10 分布式GML　出处：《江西理工大学》2012年硕士论文　论文类型：学位论文

【摘要】：GML作为一项地理信息共享标准,当前在单机环境上的存储管理理论己日趋成熟,国内先后有相关方面的研究成果,但在分布式存储并行计算环境下的理论则尚未有显著成果。分布式并行计算概念在国内外已经趋之若鹜,前几年在学术界兴起的网格计算,当今热门的云计算,以及物联网、智慧城市等,都涉及到分布式存储和并行计算处理范畴。随着知识发现与技术进步,越来越多的数据需要用空间构成3D,甚至加上时间构成4D来呈现在人们面前,这些数据是相当海量的,单机情况下的存储已经不能适应大数据量的存储处理要求,而且涉及到空间数据格式问题,为此研究分布式存储海量数据、高性能并行计算处理GML数据就是必然的。本文以单机环境GML相关存储管理理论为基础,研究分布式并行计算下GML空间数据同步复制和更新相关方面的理论,并以分布式开源数据库MongoDB为例进行实验和测试,主要包括以下4个方面： (1)分布式GML存储,提出分布式GML空间数据库内涵和3种设计方案,并以开源NoSQL数据库MongoDB为例,设计相关分布式存储方案和索引查询语言,提出云GML管理理论,并在实验中搭建分布式环境进行验证； (2)并行GML同步复制策略,对传统空间数据库和分布式空间数据库进行说明,阐述分布式数据库复制体系结构和原理,以ArcGIS10的GDB版本化同步复制为参考,提出分布式GML数据“要素分组”同步复制策略,以此处理海量GML数据在分布式集群中分发数据处理的问题,并提出冲突及解决方案； (3)分布式GML数据更新机制,分布式数据更新会造成并发和一致性问题,从分布式事务管理出发,阐述一致性更新核心技术,对ArcGIS10的长事务处理和拓扑机制进行分析,以此探讨GML几何拓扑更新一致性问题,提出分布式“GML文档化更新模型”,丰富分布式GML空间数据更新理论； (4) MongoDB GML复制更新设计,根据提出的GML同步复制和更新策略,在MongoDB搭建的分布式并行计算环境中,试验分布式GML数据同步复制和更新效率,使用分布式测试工具JMeter来测试执行情况；最终通过在分布式并行计算环境中,对GML空间数据存储管理做了底层研究,从而深入研究了其分布式数据复制和同步更新的效率机制,为GML在分布式并行计算领域开拓了一片天地。
[Abstract]:GML as a geographical information sharing standard, the current storage management theory in the single computer environment has become more and more mature, there have been related research results in China. However, the theory in distributed storage parallel computing environment has not yet achieved remarkable results. The concept of distributed parallel computing has been flocking in the world. Grid computing, which rose in academic circles a few years ago, is a hot cloud computing. As well as the Internet of things intelligent cities and so on are involved in distributed storage and parallel computing processing. With the development of knowledge discovery and technology more and more data need to use space to form 3D. Even add time to the formation of 4D to present in front of people, these data is quite massive, single-machine storage can no longer adapt to the large amount of data storage processing requirements, and involves the spatial data format problem. For this reason, it is necessary to study the distributed storage of massive data and process GML data with high performance parallel computing. Based on the theory of GML storage management in single computer environment, this paper studies the theory of synchronous replication and update of GML spatial data under distributed parallel computing. And take distributed open source database MongoDB as an example to carry on the experiment and the test, mainly includes the following four aspects: 1) distributed GML storage, the connotation and three design schemes of distributed GML spatial database are put forward, and the open source NoSQL database MongoDB is taken as an example. The related distributed storage scheme and index query language are designed, cloud GML management theory is proposed, and the distributed environment is built for verification in the experiment. 2) parallel GML synchronous replication strategy, the traditional spatial database and distributed spatial database are explained, and the architecture and principle of distributed database replication system are expounded. Taking GDB version synchronous replication of ArcGIS10 as reference, this paper proposes a synchronous replication strategy for distributed GML data "element grouping". In order to deal with the problem of distributing massive GML data in distributed cluster, the paper puts forward the conflict and solution. 3) distributed GML data update mechanism, distributed data update will cause concurrency and consistency problem, from distributed transaction management, this paper expounds the core technology of consistency update. This paper analyzes the long transaction processing and topology mechanism of ArcGIS10, discusses the consistency of GML geometric topology update, and proposes a distributed "GML document update model". Enrich the theory of spatial data updating of distributed GML; 4) MongoDB GML replication update design, according to the proposed GML synchronous replication and update strategy, in the distributed parallel computing environment built by MongoDB. The efficiency of synchronous replication and update of distributed GML data is tested, and the execution is tested using distributed testing tool JMeter. Finally, through the research of GML spatial data storage management in distributed parallel computing environment, the efficiency mechanism of distributed data replication and synchronous update is deeply studied. It opens up a world for GML in the field of distributed parallel computing.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：P208;TP338.6

【参考文献】