【作者】 陈军；

【导师】 王华军； 苗放；

【作者基本信息】 成都理工大学 ， 地球探测与信息技术， 2012， 博士

【摘要】 随着空间信息技术的发展,人们对空间信息的需求不断增加,空间分析逐渐从传统的C/S模式和B/S模式向G/S模式发展。在G/S模式下,将空间分析以服务的形式构建于云中,为行业应用提供空间分析服务,具有十分重要的研究意义。G/S模式下的空间分析服务涉及到两个关键问题,一是空间分析云服务的交换标准；二是空间分析协同计算。G/S模式以HGML为服务交换标准,但目前还处于空间数据的交换层次上,未形成完整的空间分析云服务交互标准；云计算以分布式文件系统为基础,实现了计算向存储迁移的策略,是空间分析云服务有效的解决方案。由于空间分析本身的复杂性,目前云GIS的研究和应用仍然以提供在线空间信息数据服务为主。针对目前的研究现状,本文进行了基于G/S模式的空间分析云服务关键技术研究。首先,探讨了基于G/S模式的空间分析云服务的特征,初步设计了一种空间分析云服务的基础架构。该架构由数据注册中心(G/S中心服务器、任务调度服务器集群)和存储云(空间数据服务器集群)组成。在该框架中,将空间分析云服务的网络连接分成云内连接和云外连接并采用不同的数据通讯技术以提高云服务的效率；针对空间数据的存储特征,提出了自适应间隔游程编码方法；为实现计算向存储的迁移,设计了分布式文件系统中矢量数据和栅格数据的存储结构。第二,初步建立了基于HGML的空间分析云服务交换标准。在定义了空间分析基本数据类型的基础上,给出了空间分析三个基本HGML命令GetCapabilities、 DescribeFunction和ExecuteFunction。通过这三条命令构成的标准框架,可实现所有空间分析服务的标准化表达。第三,设计并建立了G/S模式下的Map/Reduce计算框架。在分析研究了云计算的并行计算技术Map/Reduce基础上,为实现G/S模式的多种聚合模式的需要,设计了G/S模式下的Map/Reduce计算框架,并进行了数据负载均衡调度和计算负载均衡调度的深入探讨。第四,提出并建立了G/S模式下的网络地图云服务体系。针对传统B／S模式下网络地图服务的服务器瓶颈问题,建立了网络地图云服务基础架构,探讨了地图服务的Map/Reduce任务划分和任务调度方法,并通过实验验证了G/S模式下的网络地图云服务的有效性。最后,针对空间分析服务的单点聚合和多点聚合的两种聚合模式,分析了不同聚合模式下的任务调度流程,并以几何体空间操作、基于WFS的空间查询、服务器空间数据聚合模式下的栅格计算为例阐述了单点聚合的任务调度,以空间数据云下载、客户端空间插值为例阐述了多点聚合的任务调度。实验证明,本文建立的G/S模式的空间分析服务提升了空间分析服务的效率,具有一定的理论价值和实践意义。本文的创新在于以下几个方面：(1)设计了一种基于G/S模式的空间分析云服务基础架构。S端通过数据注册中心(中心服务器、任务调度服务器群)和存储云(空间数据服务器集群)的协作,具备为G端提供强大空间分析云服务能力；在该架构中,提出了一种空间分析客户端聚合服务的分类方法：单点聚合和多点聚合；为提高云端空间数据网络传输的实时性,提出了一种自适应游程编码算法,通过实验证明了算法在保证一定压缩率的同时,能实时压缩空间数据流。(2)建立了一种基于HGML的空间分析云服务交换标准。该标准中,使用三个简单的命令概括了空间分析云服务的交换方法；通过Execute命令的嵌套,使HGML具备复杂空间分析模型的表达能力。交换标准的研究为G/S模式下的空间分析云服务奠定了语言基础。(3)提出了一种G/S模式的Map/Reduce计算框架。该框架充分考虑了G/S模式中多种聚合方式,即可在客户端运行,也可在中心服务器和任务调度节点运行。由于良好的“跨端”特性,提高了空间分析模块的开发和部署效率；在该框架下,提出了一种基于处理时间的计算动态均衡调度算法,并证明了该算法的有效性；以该框架为基础,设计了一种基于G/S模式的网络地图云服务,并证明在云计算环境下,网络地图服务有效性。(4)提出了一种在分布式文件系统中空间数据存储方法。针对目前主流分分布式文件系统按字节为单位划分文件块不能有效完成空间分析分布式计算的缺点,建立了矢量和栅格数据的分块策略与完整备份策略,为空间分析计算向存储迁移奠定了数据基础。本文对空间分析云服务进行的一系列开拓性的研究和尝试,可为该领域的科研工作者、空间分析模型开发者提供理论上的借鉴,有助于G/S模式理论的进一步完善和充实,并将空间分析研究引向纵深发展。更多还原

【Abstract】 With the development of spatial information technology, people’s demand for spatial information is increasing, and spatial analysis gradually developed from traditional C/S mode and B/S mode to G/S mode. In G/S mode, spatial analysis is built on the cloud in the form of services, provides spatial analysis services for industrial applications, this has a very important significance.Spatial analysis services in G/S mode involves two key issues:one is exchange standards of the spatial analysis cloud service, and another is collaborative computing of spatial analysis. G/S model takes HGML as service exchange standards, but now it is in the level of spatial data exchange, a complete spatial analysis cloud services interact standards is not formed; cloud computing is based on distributed file system, and realizes strategy that migration from calculation to storage, it is the effective solution of space analysis cloud service. Because spatial analysis is complex, so research and applications of cloud GIS are remain online space data services. Pointing for the current status, this article researched key technology of space analysis cloud services which based on G/S model.Firstly, investigate the characteristics of space analysis cloud services based on G/S model, and preliminary design a kind of space analysis cloud services infrastructure. The architecture is consisted of data registry center (G/S center server, task scheduling server cluster) and the storage cloud (spatial data server cluster). In this framework, network connections of spatial analysis cloud services is decided into cloud internal connection and cloud outside connection, and adopt different data communication technology to improve the efficiency of cloud services; adapts one interval run coding method for spatial data storage characteristics; to realize migration from calculation to storage, design storage structure of the vector data and raster data in the distributed file system. Secondly, one exchange standard of spatial analysis cloud services based on HGML was preliminary established. Based on defining basic data types of the spatial analysis, this paper had given three basic HGML commands of space analysis: GetCapabilities, DescribeFunction and ExecuteFunction. Standard framework made of these three commands can achieve standardization expression about spatial analysis services.Thirdly, designed and established Map/Reduce computing frame based on G/S mode. Analyzed and researched cloud computing parallel computing techniques Map/Reduce, To achieve a variety of modes polymerization of the G/S mode, designed Map/Reduce computing frame work in G/S mode, deeply investigate of the data load balancing scheduling and computational load balancing scheduling.Then put forward and established network map cloud services system in the G/S mode. Pointing to server bottleneck problem of web map service in traditional B/S mode, this paper established one network map cloud service infrastructure. On the base of given network map cloud service infrastructure, investigate the tasks division and task scheduling method of map services. And experiment verified the effectiveness of network map cloud services in G/S mode.Finally, pointing to the two aggregation modes which are the single point aggregation and multi polymerization spatial analysis mode, analysis task scheduling process of different polymerization mode. And elaborate task scheduling of single point aggregation according to space operations of geometry, spatial query based on WFS, grid computing in the server space data aggregation mode. Elaborate multi polymerization task scheduling technology according to spatial data cloud download, spatial interpolation of the client. Experiment results show that spatial analysis service based on G/S mode which established in this paper have enhanced the efficiency of spatial analysis service, and have a certain theoretical value and practical significance.The innovations of this paper are the following aspects:(1) Design infrastructure of space analysis cloud services based on G/S mode. The S end provide powerful spatial analysis cloud service capabilities for G end through the registered data center (central server, the task scheduling server farm) and the storage cloud (spatial data server cluster) collaboration; in this architecture, a spatial analysis client aggregation service classification is put forward:a single point aggregation and multi-point polymerization; presents an adaptive run-length encoding algorithm to improve the real-time nature of the cloud spatial data network transmission. Experiments show that the algorithm can guarantee compression rate while real-time compress spatial data streams.(2) Establish one space analysis cloud services exchange standards based on HGML, which rich the theoretical content demonstrated by G-end space data. Summarizes the spatial analysis cloud services exchange method by three simple commands; Execute command nesting making HGML have expression skills of complex spatial analysis model. Researching on exchange standards laid the language foundation for space analysis cloud service in G/S mode.(3) Proposed one Map/Reduce computing framework in G/S mode. The framework fully consider the multi polymerization manner in G/S mode, and can be run on the client, and can also be run in the central server and task scheduling node. It has Cross-end features, and improves the development and deployment efficient of spatial analysis module; in the framework, processing calculated dynamically balanced scheduling algorithm based on operation time, and proved the effectiveness of the algorithm; taking framework as the basis to design a network map cloud services based on G/S model, and proved the effectiveness of the Web Map Service in cloud computing environment.(4) Put forwarded one spatial data storage method in distributed file system. At the present distributed file system divided file block to sub-units by bytes, and it can’t effectively complete space analysis distribute calculate. So this paper built partitioning strategy and complete backup strategy for vector and raster data. Laid the data foundation for space analysis migration from calculate to storage.In this paper, a series of pioneering research and try of the spatial analysis are conducted. It can provide a theoretical reference for scientists in cloud services field and space analysis model developers, contribute to further improve and enrich of the G/S mode theory, and lead spatial analysis to depth development.更多还原