【作者】 李源林；

【导师】 谢忠；

【作者基本信息】 中国地质大学 ， 地图制图学与地理信息工程， 2013， 博士

【摘要】 地理信息系统(GIS)与互联网的结合,拓展了GIS的新领域和新途径,极大的促进了地理空间信息的应用推广,让跨地区和跨行业的空间信息共享更加方便,也使基于地理信息的大众化应用得以快速发展。计算机硬件性能的快速提高,软件领域新构架新算法的推出,以及GIS学科的不断进步,都推动了网络GIS的快速发展。网络GIS应用的不断深入,带来了一些新的问题和挑战,表现有：互联网巨大的用户量和GIS海量数据给网络GIS应用带来很大的性能压力；软件落后于硬件发展的现状,使得新硬件对网络GIS系统性能提升作用受限；传统的网络GIS软件和开发模型不足以满足GIS应用快速开发的需求。网络GIS集群继承了计算机集群的特点,保证了服务和应用的稳定,极大的缓解了服务端压力,同时方便扩展。负载均衡策略和算法是集群的一项关键技术,算法的改进对提升集群性能和稳定性意义重大,也是各研究领域包括网络GIS的研究热点。服务器虚拟化技术提供了建设集群的新方法,利用它可以很便利的使用各类服务器和PC来构建虚拟集群,已经有不少学者和GIS厂商的研究人员提出了构建虚拟化网络GIS集群的方法和模型,各自有其特色和不足。本文结合国家科技支撑计划项目“地理空间信息工具集网络服务平台研发”,主要围绕基于服务器虚拟化的网络GIS集群的应用模型和关键技术的实现方法,重点研究网络GIS集群模型的构建方法、虚拟资源的分配技术、负载均衡调度策略和算法、服务与接口模型等一系列关键问题。在此基础上设计结构合理的网络GIS集群模型,解决集群中虚拟资源的动态分配问题,实现符合GIS应用特点的负载均衡算法,设计简明实用的网络GIS服务与接口模型。本文具体的研究工作如下：(1)研究分析了服务器虚拟化技术、网络GIS以及集群技术的研究现状和发展趋势,指出虚拟化和集群在提高硬件资源利用率、系统性能和可用性等方面具有独特的优势,是提高网络GIS集群可靠性和性能的有效技术手段。同时虚拟化和集群发展中存在着一些问题和难点,例如性能损耗、负载调度等,都成为GIS集群优化工作所要研究和解决的问题。(2)研究了虚拟化网络GIS集群结构模型。在研究服务器虚拟化的特点和实现的技术层次基础上,指出虚拟化技术是提高服务器硬件尤其是处理器利用率的有效手段,在基于多核处理器的微小型服务器迅速发展的当前,构建于廉价服务器之上的集群系统能够处理时空复杂度高的计算密集型问题,并可应用于GIS领域。以国内外主流GIS平台提供商的集群结构为研究目标做了分析和对比,结合服务器虚拟化技术,提出了一种网络GIS集群结构模型,该模型采用处理器核心、操作系统、GIS服务实例数量之间的1：1：1映射关系建立集群,并用实验测试的方法将其与单物理服务器多GIS服务实例的方案进行对比。(3)研究了虚拟化网络GIS集群资源分配技术。分析了虚拟机资源分配的一般方法和虚拟机放置策略,提出了一种基于集群负载预测的资源动态分配技术。对集群的中长期历史负载进行了时间序列建模,并以一组气象数据为实验对象进行了预测建模实验,在负载预测的基础上设计了虚拟机放置时机决策的策略和算法,以及基于最少优先策略的虚拟机放置目标选择算法,并描述了上述算法流程和伪代码。设计了基于分配算法的实验场景并在气象应用上进行测试,对分配技术进行验证。(4)研究了虚拟化网络GIS集群负载均衡算法。首先研究了遗传算法特点,以及它在负载均衡调度上的应用情况,表明它对网络GIS负载均衡调度同样具备适应性和和鲁棒性。根据GIS运算和应用的特殊性和复杂性,建立了基于任务优先级和执行时间的GIS任务模型,作为遗传算法基因编码依据,设计了用于负载均衡的遗传算法,包括适应度函数、遗传算子以及控制参数,并形成了算法流程。最后将算法插入网络GIS集群负载均衡模块进行了实验测试。(5)研究了虚拟化网络GIS集群服务与接口模型。在研究空间数据共享问题和面向服务构架的问题基础上,针对SOA结构厚重,数据模型复杂的不足,提出了基于ROA的轻量级GIS集群服务与接口模型,以及基于RIA的跨浏览器的客户端模型,降低了接口模型复杂度,方便用户使用。(6)研究了虚拟化网络GIS集群服务原型系统。设计并建立了基于服务器虚拟化技术的网络GIS集群原型框架,从具体技术实现的角度进行了原型的分析与描述。以气象行业某应用为研究和实验对象,将虚拟化网络GIS集群原型付诸实践,并进行性能测试,验证集群原型的可行性和合理性。更多还原

【Abstract】 The combination of GIS and Internet has expanded new fields and new channels of GIS, and promoted the application and extension of geographic spatial information dramatically, which not only makes it more convenient to share the spatial information cross-regionally and cross-industrially, but also makes it more rapidly to develop the popularization of geographic information. The rapid development of WebGIS has been pushed forward owing to the great enhancement of the computer hardware performance, the promotion of new frameworks and new algorisms in software field, and continuous improvement of GIS subject. The unceasing penetration of the WebGIS application has brought about some new problems and challenges, for instance:the large numbers of Internet users and the mass data of GIS have made GIS application under great performance pressure; since software development lags behind hardware development at present, the role that new hardware plays in promoting WebGIS performance has been limited; the traditional WebGIS software and development model couldn’t satisfy the demand of GIS rapid development.WebGIS cluster has inherited the characteristics of computer cluster, which ensure the stability of service and application, release the server pressure dramatically, and make it more convenient to extend. Load balancing strategies and algorithms are one of the key technologies of cluster. The improvement of algorism is essential to promote cluster performance and stability, which is also the research focus of various research fields including WebGIS. The server virtualization, which has provided new methods for cluster construction, can be used to establish virtual cluster based on different servers and PCs much easily. Many scholars and researchers from GIS industries have put forward new methods and models to establish virtual WebGIS despite the fact that they all have their own advantages and limitations.On the basis of the project supported by the national science and technology, this paper focuses on the WebGIS cluster application models and the key technology realization methods based on the server virtualization, lays emphasis on such important issues as WebGIS cluster model construction, virtual resource allocation technology, load balancing strategies and algorisms, and service and interface models and so on. On this basis, the structurally proper WebGIS cluster model has been designed; the dynamic allocation problem of virtual resource in cluster has been solved; the load balancing algorism according with the features of GIS application has been realized; and the simple and applicable WebGIS service and interface model has been designed.The details of this study are stated as follows:(1) In this study, the present research conditions and tendencies of the server virtualization, WebGIS and cluster technology have been analyzed, which concludes the advantages of virtualization and cluster in promoting hardware resource utilization rate, system performance and usability and so on. Therefore, it makes virtualization and cluster effective technical means in enhancing WebGIS cluster reliability and performance. At the same time, some problems and difficulties in virtualization and cluster development, such as performance loss and load scheduling and so forth, must be studied and solved.(2) This paper then studies the structure model of WebGIS virtualization cluster. On the basis of studying the features and realization technologies of server virtualization, virtualization technology has turned out to be an effective method to improve the utilization rate of the server hardware, especially that of the processor. Under the current circumstances that the micro server based on multi-core processor has gained rapid development, the cluster system built on cheap server can deal with computational density problems of high time-space complexity, and can also be applied in GIS field. Analysis and comparison have been made between the cluster structures developed differently by national and international mainstream GIS platform providers. A structure model of WebGIS cluster has been proposed in connection with server virtualization technology. This model constructs cluster with1:1:1mapping among the numbers of processor core, OS, and GIS service instance. It has been compared with the solution of singular physical server with multi GIS service instance by experimental tests.(3) This paper also studies the resource allocation technology of WebGIS virtualization cluster. By analyzing ordinary methods of virtual machine resource allocation and placement strategies of virtual machine, a dynamic allocation technology of recourse based on cluster load prediction has been proposed. The time series modeling of the medium and long term cluster load history has been realized. A set of meteorological data has been experimented to test and verify this modeling. Strategies and algorisms of the virtual machine placement timing have been designed on the basis of load prediction. The algorism of virtual machine placement target selection has been put forward based on the least first strategy. The above algorism procedure and false code have been described. The experimental scene based on allocation algorism has been designed and tested on meteorological application, and the allocation technology has been verified.(4) The load balancing algorism of WebGIS virtualization cluster has been studied too. The features of genetic algorism and its application to load balancing scheduling have been studied at first, which indicates that it is adaptable and robust in WebGIS load balancing scheduling. According to the particularity and complexity of GIS algorism and application, the GIS task model has been constructed as the basis for genetic coding of genetic algorism based on the task priority and execution time. The genetic algorism of load balancing, including the fitness function, genetic operator and control parameter, has been designed, and algorism procedure has been formed. At last the algorism has been inserted in the load balancing module of WebGIS cluster to be experimented and tested.(5) The service and interface model of WebGIS virtualization cluster has been studied further. On the basis of studying the problems of the spatial data sharing and service frameworks, the service and interface model of lightweight GIS cluster based on ROA and the cross-browser client model based on RIA have been put forward to solve the problems of heavy SOA structure and complicated data model, lowering the complexity of interface model so that the customers can use it at ease.(6) At last the service prototype system of WebGIS virtualization cluster has been studied. The WebGIS cluster prototype framework based on server virtualization technology has been designed and constructed. The analysis and description of this prototype have been made from the aspect of specific technology accomplishment. Certain application in meteorological industry has been studied and experimented in order to put WebGIS virtualization cluster prototype into practice, make performance tests and verify the feasibility and rationality of cluster prototype.更多还原