【作者】 张卫；

【导师】 查亚兵；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2009， 博士

【摘要】 大规模并行分布式仿真需要大量的资源和多个组织结构间的协作,网格能够实现跨组织机构的资源共享与协同问题求解,它为大规模并行分布式仿真的开展提供了有效的解决途径。当前网格技术正朝着面向服务架构方向发展,通过服务为网格用户提供各种功能,以确保网格功能模块的松耦合性和在异构环境中的互操作能力。为了支撑大规模并行分布式仿真的开展,可以在网格环境中提供一组标准的服务,通过构建面向并行分布式仿真服务网格来使能仿真相关资源的共享和协同仿真的开展。首先给出了面向并行分布式仿真服务网格的使用想定,基于给出的使用想定进行了需求分析,确定了面向并行分布式仿真服务网格应该提供的若干关键服务和功能,主要包括HLA RTI服务、资源发现服务、仿真执行管理服务和仿真任务迁移等。接着,在研究了相关技术的基础上,确定面向并行分布式仿真服务网格的总体框架,明确了面向并行分布式仿真服务网格的层次结构,这些层次包括资源层、仿真基础服务层、仿真核心服务层、仿真门户层和用户层。为了在面向并行分布式仿真服务网格环境中将HLA RTI作为基础仿真服务提供,需要实现HLA RTI的Web服务化。在分析了HLA和Web服务在军事领域和商业领域上下文互操作性异同的基础上,给出了HLA RTI Web服务化概念和基本方法,重点讨论了HLA RTI Web服务化需要解决的关键技术问题和可能的解决方案,关键技术问题有HLA RTI Web服务接口定义、消息编码方式、调用状态维护、部署模式、数据交换模式调和等,在解决了这些关键技术问题的基础上,提出了一种用于实现HLA RTI的Web服务化方法,并给出了基于Globus工具箱的实现。此外,讨论了联邦成员Web服务化的相关问题和方法。面向并行分布式仿真服务网格需要通过资源发现服务使能服务网格中的资源发现。资源发现以资源描述为前提,提出了一种基于语义的资源描述方法,给出了相应的资源匹配方法。该资源描述方法通过建立相关领域的本体,在资源描述过程中使用本体定义的相关概念和属性,为资源描述引入了语义,改进了资源描述缺乏语义而导致的资源匹配准确性差和完备性不高等问题。在建立了基于语义的资源描述方法之后,给出了一种具有自组织能力的资源发现服务,通过自组织能力简化了服务网格的管理和使用、克服了集中式资源发现服务所固有的性能瓶颈和单一故障点问题。服务质量保证是网格技术区别于其它分布式计算技术的重要标志之一,服务网格必须为并行分布式仿真系统在网格环境中的运行提供负载平衡机制,而仿真任务迁移是实现负载平衡的主要手段之一。在研究了一般性迁移问题的基础上,重点分析了基于HLA分布式仿真中联邦成员可以采取的迁移方法和协议,提出了一种基于移动代理的成员迁移机制,实现了并行分布式仿真系统在网格环境中的负载平衡,改进了并行分布式仿真在网格环境中的执行效率。计算资源是服务网格中的一类重要资源,也是执行大规模并行分布式仿真最为需要的资源,面向并行分布式仿真服务网格必须为并行分布式仿真作业在这些大量计算资源上的执行提供支持。在研究了仿真作业管理与调度方法的基础上,为并行分布式仿真在服务网格环境中的执行设计了仿真执行管理服务,作为对仿真执行管理服务的支撑,设计了仿真基础执行服务,讨论了如何将本地计算资源封装为仿真基础执行服务。在以上对面向并行分布式仿真服务网格总体框架和关键技术研究的基础上,设计并实现了原型系统,通过以XX对抗仿真系统为原型的仿真实例验证了面向并行分布式仿真服务网格关键技术研究工作的可行性,实现了仿真资源的共享,支撑了协同仿真的开展。更多还原

【Abstract】 Large scale parallel and distributed simulations require a multitude of resources and collaboration between simulation participants. Grid can enable resource sharing and collaborative problem solving across multiple organizations, so it provides an efficient way for large scale parallel and distributed simulations. And currently, grid technology is developing towards Service Oriented Architecture, and provides varieties of functions to grid users through services in order to guarantee the loose coupling of functional modules of grid and their interoperability in heterogeneous environments. To support the fulfillment of large scale parallel and distributed simulations, it is possible to construct the service grid for parallel and distributed simulations by providing a set of standard services, ultimately to enable the sharing of simulation relevant resources and the achievement of collaborative simulations.The dissertation firstly described the application scenario of the service grid for parallel and distributed simulations, and identified the key services and functions that the service grid must provide, through demand analysis. These key services and functions included High Level Architecture(HLA) Runtime Infrastructure(RTI) service, resource discovery service, simulation execution service, and simulation task migration. Then on the basis of the research on relevant technologies, the overall framework of the service grid for parallel and distributed simulations was determined. It clarified the hierarchy of the service grid, which was composed of resource level, simulation basic service level, simulation core service level, simulation portal level, as well as user level.Service enabling of HLA RTI was indispensable to provide RTI service in the service grid. The difference of interoperability between HLA in military domain and Web service in commercial domain was analyzed. And afterwards, the concept of HLA RTI’s service enabling and its concrete methods were also given. As an emphasis, the key problem and possible solutions of service enabling of HLA RTI were discussed, in terms of the definition of Web Service interface for HLA RTI, message encoding and decoding, invoking status maintenance, RTI deploy pattern, etc. Based on the solutions of these problems, an approach to web enabling HLA RTI and its implementation based on Globus Toolkit were given. In addition, the problem of federates’web service enabling and its solutions were discussed.The discovery service was indispensable for the service grid to enable resource discovery in grid environment. The resource description was prerequisite for the resource discovery. This dissertation proposed a semantics-based resource description method and corresponding resource matching methods. This method constructed relevant domain ontology for resource description at first, and then used the concepts and relations defined in this ontology to describe the resource; therefore the semantics were introduced for the resource description. As a result, inaccuracy and incompleteness that resulted from semantics shortage were fixed. After constructing the resource description, a discovery service with the ability of auto-organization was proposed. It facilitated management and usage of the service grid, and surmounted performance neck and single fault point that implicitly existed in centralized resource discovery systems.Service quality assurance was one of the main characteristics that distinguished the Grid technology from other distributed computing technologies. The service grid needed to provide load balance mechanism for parallel and distributed simulation’s execution in grid environments. Simulation task migration was one of main measures to achieve load balance. The studies clarified general migration issue, emphasized on migration protocols and methods for HLA based simulations, and proposed a federate migration method based on mobile agent. And eventually, the load of HLA based simulations was balanced, and promoted the execution efficiency of parallel and distributed simulation in grid environment, through the said method.Computing resources was one kind of most important resources in the service grid, and was also the most requisite for large scale parallel and distributed simulations. The service grid must support the job of large scale parallel and distributed simulations to run on these multitudes of computing resources. The management and scheduling algorithm of simulation jobs were researched. Based on the research, a simulation execution management service was designed as one important part of the service grid. As a supplement for the simulation execution management service, the simulation basic execution service was designed and further it is discussed how to encapsulate local computing resource as simulation basic execution service.Based on the proposed overall framework and key technology, a prototype system was designed and implemented. And significantly, it was validated that the above-mentioned key technologies were feasible to fulfill the sharing of simulation resources and supporting the achievement of collaborative simulation, under a typical scenario of some simulations confront system.更多还原