【作者】 马卫东；

【导师】 周明天； 李幼平；

【作者基本信息】 电子科技大学 ， 计算机应用技术， 2007， 博士

【摘要】 信息的理解与共享是信息科学中十分重要的课题。Internet及其上的信息服务系统无疑是最重要的信息共享基础设施,数以亿计的用户和各类软硬件参与到了这个复杂巨系统的发展与演化之中,使其成为规模宏大的复杂网络。Internet是一个巨大的信息仓库,是一个庞大而混沌的网络。Internet信息的浩瀚使得信息的获取与查找变得相当困难。如何让Internet提供高效、方便、及时、准确的信息服务成为当前研究的热点。实证研究表明,由Internet所构成的复杂巨系统在整体规模上呈现出自组织、小世界和幂律分布等许多新的特征。群体用户对网络资源的需求也应当存在某种整体特性。通过大量实验数据的收集,发现群体用户对资源的需求具有高度的同一性特征,实证了用户需求二分网络的幂律分布规律。用户需求的幂律分布特点表明了存在一个高热度的信息服务小子集,能够满足大多数用户的信息服务需求。将这个高热度的信息服务小子集作为信源进行主动、及时的信息分发,可以改善Internet网络的流量分布,使其更高效的工作。这种根据用户的需求,将其所需要的信息及时、准确地送到他们身边的信息服务技术,称之为主动服务。论文整体上分为两大部分。第一部分,研究了Internet信息共享动力学问题,通过实证和仿真分析探讨了用户需求行为对互联网整体特性的影响。该部分内容成为信息共享主动服务技术的理论基础。第二部分,研究了主动服务体系结构、网络层次模型和实现关键技术等问题。两部分研究内容既相关又有所不同,相辅相成,完整的对信息共享主动服务理论提供有效的支撑。论文的主要贡献和创新点如下:1)首先对Internet用户访问需求进行了实证研究。实验数据表明,群体用户的需求具有高度的共性特征,通过区域用户需求行为的统计特性,发现了用户需求二分网络的幂律分布规律,其幂指数介于1.7到1.8之间。讨论了该特征对互联网整体特性的影响。建立了虚拟资源网和物理拓扑网络双层模型,仿真研究表明幂律分布的用户行为会恶化Internet网络行为。将高幂律小子集通过分布式映射存储,能够大大改善网络性能,使得相变节点右移,为主动服务奠定理论基础。2)提出了主动服务体系结构及网络层次模型,详细讨论了一个基于Web服务的信息订阅/发布主动服务系统的实现模型、工程框架和关键技术。该模型基于Web服务技术建立一个互联网海量信息自动获取与分发系统,为用户提供高效、及时的、个性化的按需服务。3)研究了主动服务需要的网络链路支撑技术。包括数字视频广播DVB标准和可用的互联网链路技术。基于DVB的信息主动服务是一个高效的、非常具有应用前景的系统。提出并完成了一个基于Linux操作系统的IP/DVB网关的设计与实现。4)从信息查找、检索的服务质量要求出发,探讨了主动服务涉及的语义信息处理关键技术途径和实现手段。包括语义信息处理的数学模型、聚类和分类技术等信息与知识获取的研究进展情况。5)提出了一个面向互联网信息获取与分发系统的统一内容标识与定位(UCI/UCL)概念、模型和实现方法。包括UCI本体库的构建、UCI标引的自动生成、UCI/UCL的分层模型、UCL多层次映射及UCI/UCL在互联网海量信息自动获取与分发系统中的应用与实验情况。6)探讨了面向主动服务的数据分发服务质量特性、数据广播调度算法。包括基于Push的主动广播调度、基于Pull的按需广播调度和两者结合的混合广播调度,最后提出了一个面向主动服务系统的自适应混合广播调度α-β-γ模型。更多还原

【Abstract】 The millions of information service systems based on Internet are composed of complex networks. All of the people and computer software take part in the evolution and development of the huge complex system. Internet is not only a huge information library, but also a chaos network. Therefore the interested information by the users is extracted and found difficultly. This dissertation want to propose a model called active services to implement the efficient, conveniency, exactly and real-time data dissemination based on Internet and data broadcasting system.First, this dissertation discusses the influences of user requirement behaviors to Internet collective dynamics. The dynamics of the complex networks such as Internet that is mainly influence by user requirement behaviors, and can be statistic in time series. A large number of complex networks, both natural and artificial, share the presence of highly heterogeneous, scale-free degree distributions and small-world phenomena. This dissertation analyzed the empirical collective behaviors of user requirements in a region, discovered that the frequency and the in-degree distributions of bipartite networks constructed by user visiting web sites are power-law, and the exponential is between 1.7 and 1.8. A novel two-tier model, the virtual resource networks (VRN) and physics topology networks (PTN), is proposed to study the influences about the Internet collective behaviors. The mapping process mechanics of VRN to PTN simulates that the dynamic behaviors of resource network how to influence the internet physics topology networks. The power-law characteristics of VRN can bring forth the results that the phase transition critical point moves left and network performance is more incapable. The distributed mapping process of VRN to PTN for the small subset with the high degree nodes was constructed, then the phase transition critical point of PTN moves right and the network performance is improving obviously.Second, this dissertation submits a kind of active service architecture and its network model, and discusses a publish/subscribe system model based Web services, including implementation frameworks and key technology. This model wants to establish a huge of information extraction and dissemination system for the Internet and other data broadcasting system, so that supports efficient, personalized, and timeliness service-on-demand.In the chapters 4 to chapter 7, the dissertation discusses many key technologies for active services. The active services are supported by many network technologies, including DVB standard, IP multicast, and active networks etc. A design of IP/DVB gateway device based on Linux is submited int chapter 4. The information sharing semantic processing, the QoS of information search, the semantic Web, the information clustering and classification are analyzed. In the chapter 6, the models of Uniform content identifier (UCI) and uniform content locator (UCL) are submitted and have been discussed for concepts, methods, and implementations, including construction of ontology library, automotive indexing for UCI/UCL, the network level model and mapping mechanics, and the experimentation and application of UCI/UCL.The data dissemination and dynamic broadcasting scheduling algorithms are analyzed in the last chapter. The asymmetric communication system, where the server has much communication capacity than clients, is emerged in modern network environment. The data dissemination and data broadcasting are an efficient way to transmit data. Broadcast can be either pull-based, where clients ask for data from the server, or push-based, where the server pushes data to clients. The broadcast schedule is important to the overall system performance. In the chapter, the scheduling algorithms, performance metrics and characteristics of push-based active broadcasting and pull-based on-demand broadcasting are discussed. An adaptiveα-β-γhybrid broadcasting scheduling model is submitted for active services environment.更多还原