【作者】 王道谊；

【导师】 刘元安；

【作者基本信息】 北京邮电大学 ， 电磁场与微波技术， 2010， 博士

【摘要】 本文以多业务系统融合相关通信技术和信息技术为出发点,论述了多业务系统融合中内容分发技术和服务交付平台技术的研究现状和意义,综合考虑多业务系统融合环境中网络资源和IT资源动态重构、商业运营流程优化与变更、客户服务优化、多业务扩展和内容与应用供应商合作支撑等方面的需求,具体研究开发了服务交付平台所需的多项关键技术,并深入研究了内容流行度分布特性和内容分发策略。论文主要创新成果如下。1)研究了信息通信服务的属性数据抽象和服务提供过程的动态特性,提出了信息通信服务框架模型的IT架构视图、数据共享视图和三维视图。该框架模型在水平方向给出了运营商业务平台的IT架构和数据共享架构,在垂直方向给出了产业链中各方在服务提供过程中所需的服务管理数据支持和信息通信支持。该框架模型是研究多业务系统功能需求的关键理论模型。2)提出了内容流行度、内容流行集中度、内容流行百分点集中度的规范定义；提出了内容流行度分布集中性的定量描述方法,包括Lorenz曲线、Gini系数、差异系数和Theil系数。利用Zipf分布仿真验证了这些方法的有效性,给出了Zipf分布参数拟合的数学方法,提出了基于集中性参数的近似估计算法。这些方法可以为研究内容集合访问特性、制定内容分发策略、进行内容分发系统设计与仿真提供理论指导和实用工具。3)提出了体现内容流行度集中性分布的节目访问模型的构建方法,开发了包括节目模型、网络模型、业务访问规则和仿真算法在内的系统仿真模型。在此基础之上,仿真研究了本地缓存内容形成规律以及核心节点处理能力、边缘节点缓存能力对系统服务质量的影响。研究了基于“最近访问”和“最为流行”的边缘缓存策略之间的相关性以及内容流行度分布特性对它们的影响,并利用Zipf分布进行了仿真验证。这些研究工作可以用于构建内容分发系统基础研究环境。4)提出了基于时间窗的内容更新缓存算法,并构建节目更新模型对该算法进行了仿真验证。通过该方法,边缘缓存可以快速识别新内容的流行度级别,并且不会丢失对原有缓存内容的流行度知识积累,从而可以快速处理内容更新给系统带来的影响。5)提出了基于组播单播自适应调度的突发访问算法,并构建突发访问模型对该算法进行了仿真验证。该方法通过自动探测、动态调度单播资源和组播资源,实现突发热点内容的快速发现和资源自动分配,从而避免突发访问造成的系统性能大幅度下降。该算法尤其适用于有线数字电视系统和双向IP网混合组网的融合环境,有较好的实用性。6)在分析研究服务交付平台功能需求和技术需求的基础上,提出了服务交付平台的逻辑功能架构和物理实体架构,开发了设备接入认证、服务发布和服务接入一站式认证等协议,完成了相关功能实体组件关系模型、消息交互规程和关键Web服务的定义。完成了终端管理实现方案,并针对有线电视网络数字机顶盒给出了参数扩展方法。完成了业务无关的网络接入控制方案。该服务交付平台从客户服务、运营商内部商业流程优化、运营商网络资源和IT资源动态重构等方面支持对原有业务系统的融合和新业务的快速开发,具有非常好的可扩展性。更多还原

【Abstract】 Based on the latest technologies on convergence of telecommunication and information industry, the content delivery and service delivery related technologies for multi-services convergence were discussed in this thesis. By taking into account operator’s network resources and IT resources dynamic reprogramming, customer service optimization, operator’s business processes improvement, multi-services scalability, and content/application suppliers’ cooperation requirements, several key technologies related to Services Delivery Platform (SDP) were developed. Moreover, the content popularity characters and the content delivery polices were studied. The contributions and achievements are as follows.1) Based on the analysis of the attributes of service and service operations, a service framework model was proposed. The service framework model are defined from three points of views, i.e. the IT architecture view, the data storage view and the global view. The model indicates how to architect the common functionalities among various services in the horizontal direction, and how to delivery control, management and media data in the vertical direction.2) The definition of content objects’popularity, concentrated-popularity and percentile-popularity were put forward. Several methods for measurement of the inequality of objects’popularity were developed, including Lorenz curve, Gini coefficient, Coefficient of Variation, entropy and Theil index. The results of the simulation on Zipf-like distribution show that these definitions can better describe the content objects’popularity and its inequality. At the same time, the index parameter estimation algorithms of Zipf-like distribution were also put forward.3) The popularity-based content access model was brought forth, and a general system simulation model was developed, including content object model, network grid architecture, service request process rule, and simulation algorithm. Through system simulation, the effects how content delivery network works, and the effects of the core node throughput capability and edge node caching capability were studied. Then, the relationship betweent latest-accessed-based and the most-popular-based caching algorithms were studied. The given simulation model is an useful platform for content delivery mechanism research, and the simulation results indicate several key aspects related to content delivery network deployment.4) A movie-access-model with content updating was built, and an improved time-window-based content caching algorithm was proposed to solve the problem caused by content updating. The movie-access-model and the caching algorithm were investigated through network simulation. The simulation results show that the given algorithm can cope with the challenge caused by content updating.5) A flash crowd access model was built, and a multicast-unicast adaptive algorithm for flash crowd alleviation was put forward. The simulation results show that the adaptive algorithm can detect flash crowd conditions automatically, and assign multicast resources dynamically to alleviate the traffic burst.6) The functional entities architecture and physical entities architecture of SDP were developed. Based on the given architecture, several key protocols for common functionalities were proposed. These protocols include customer device access authentication and key resources later-binding, service go-to-market process, service single-sign-on, customer device congfiguration and management, network access control and automatic configuration, etc. The component relation model, interoperation model and Web service description of the key functional SDP entities were deployed. The given architecture can maintain its scalability though dynamic re-using and later-binding the capabilties provided by operators’network resources and IT resources.更多还原