【作者】 张伟；

【导师】 李三立；

【作者基本信息】 清华大学 ， 计算机科学与技术， 2011， 博士

【摘要】 移动互联网和移动视频网格越来越广泛地应用于包括视频直播、视频点播、视频监控等视频流媒体服务领域。移动视频网格存在异构、终端移动、无稳定拓扑结构等特点，无线网络带宽仍受很大局限，移动节点间还有带宽竞争；而移动网格中视频资源访问相比其它传统网格服务要求更大更稳定的网格带宽资源。为解决这个矛盾，利用网格中节点的协作，在移动节点访问视频资源时提供质量控制，保证视频资源访问路径的实际带宽满足视频传输要求，保证网格中节点移动时也能使访问视频资源更平滑，是本文研究的主要问题。移动服务代理在移动网格中位置关键，移动节点到移动服务代理间的访问路径对网格视频资源访问路径有重要影响，所以本文将重点从移动视频网格链路质量控制，移动服务代理访问路径及使用模式等方面，研究移动视频网格中的资源访问路径的发现与管理。本文的主要工作包括：（1）提出支持视频资源访问质量控制的资源访问路径发现方法Q-RAPD，可以保证每个被准入的视频资源访问的路径带宽满足视频流码率要求，且单路径模式的实际容量可以达到理论优化值。（2）提出移动视频网格资源冗余访问路径，研究了在冗余路径中同时使用多个服务代理支持同一个资源访问的数据编码传输模式，与只使用单条路径相比，冗余路径方案可以为资源访问获得更稳定的网络带宽。（3）在Q-RAPD与冗余路径研究基础上提出支持视频资源访问服务质量控制的资源访问多路径发现与平滑切换方案Q-RAMPDS，其中多条资源访问路径承载的带宽份额可根据被服务节点移动状态自适应变化，保证了在节点移动中视频资源访问仍能保持平滑，网络带宽容量也可保持在较高水平。（4）提出了移动网格中间件开发框架MG-MDF，通过在待开发协议或网格中间件功能模块与操作系统之间构造一个对于运行环境可适配的中间层，可以有效降低异构网格中开发调试的复杂度，保证不同运行环境下开发的一致性。（5）以上述理论及实验研究为基础，搭建了一个移动视频网格实验床，由集群式流媒体服务器，移动网关，无线摄像机，以及其它有线及无线节点组成，除已承载的视频监控服务外，它还可以支持其它的网格视频服务，是一个有扩展性的平台。更多还原

【Abstract】 Mobile Internet and mobile video grid is increasingly widely used in the scope ofvideo streaming services, including video broadcast, video on demand, videosurveillance and other services. The network of mobile video grid has the followingunique features: hetorogenity, terminal mobility and unstable topology; the wirelessnetwork bandwidth still has its significant limitation, so that there are competitionsbetween mobile nodes. However, the video resource access in mobile video gridrequires much more stable bandwidth than the other traditional grid services. To solvethe contradiction, collaborating with the nodes in the grid, supporting quality controlwhen the node accesses video resources, ensuring the actual available bandwidth of theresource access path to satisfy the transmission requirement of the accessed video, andenabling to access the video resources more smoothly when the node is moving, theseare the main research subjects of this paper. Mobile service agent is of key position inthe mobile grid. The path from mobile node to mobile service agent has very essentialimpact on the resource access path. Therefore, this paper studies the discovery andmanagement of resource access path in mobile video grid from three aspects: qualitycontrol of the link in mobile video grid, access path of the mobile service agent and theservice mode of resource access path.The main work of this paper includes:(1) Q-RAPD (QoS-Resource Access Path Discovery) in mobile video grid isproposed. It ensure that during every accessed video resource the request bandwidth ofthe link can satisfy the video bitrate, and the actural network capacity in single-pathmode can reach the range of the theoretical optimum value.(2) Resource access redundant paths in mobile video grid are proposed. It isstudied with using multiple mobile service agents simultaneously to access the samedata in redundant paths and the encoding transmission mode. Compared to single-pathmode, redundant paths can achieve more stable network bandwidth for resource access.(3) Q-RAMPDS (QoS-Resource Access Multiple Path Discovery and Switching) isproposed based on the study of Q-RAPD and redundant paths. In Q-RAMPDS the ratioof every resource access path’s bandwidth can change adaptively according to the moving status of the mobile node served. It ensures that the resource access whilemobile node is moving can be smooth, and the network capacity can be maintained at ahigh level.(4) MG-MDF (Mobile Grid Middleware Development Framework) is proposed.The main idea is to construct a middle layer between the reality of the protocol or thegrid middleware module and the operating system or the simulated system. It canreduce the complexity in the mobile grid development, and ensure the consistency ofthe development under different run time.(5) Based on the above studies of theories and experiments, a mobile video gridtestbed is built. The testbed is constructed by cluster streaming server, mobile gateways,wireless cameras and other wired or wireless nodes. In addition to implement the videosurveillance services, it can expand to support other grid video services. It is a scalableplatform.更多还原