Efficient and Scalable Streaming Techniques for Large-Scale Delivery of Stored VBR Video

许多网络应用，如电子商务、远程教育、数字图书馆、多媒体电话会议和在线娱乐，都涉及跨异类网际网络将存储的视频实时传送到大量客户。为了在客户端连续回放，必须以端到端的方式提供服务质量(Qos)。高效的大规模视频传输系统的设计因高度突发的可变比特率(VBR)压缩视频、底层的异质网络环境、不同的客户端功能以及不同的客户端服务质量要求等挑战性问题而变得复杂。解决这些问题的解决方案必须既高效又可伸缩。拟议研究的主要目标是设计和开发高效且可伸缩的流媒体技术，用于大规模传输存储的VBR视频。该项目依赖于一种新颖的基于分发树的视频交付系统架构，该架构使用一组战略性地放置在网际网络中的代理服务器来执行全网络范围的视频流和分布式媒体控制。基于这种体系结构，研究人员计划开发和研究高效且可扩展的实时视频流机制，该机制能够有效地适应底层网络环境和客户端系统的异构性，同时为大量客户群提供灵活的Qos支持。特别是，研究人员将开发网络范围的视频平滑技术，该技术利用代理服务器和客户端的内存缓冲区来减少视频传输中的速率变异性，并适应网络带宽限制。O研究人员将开发代理视频分段技术，该技术利用网络边缘代理服务器的磁盘存储空间，在主干广域网(广域网)中实现显著的带宽削减。O研究人员将开发自适应速率和差错控制机制，如选择性帧丢弃和选择性重传，这些机制利用特定于应用的信息来满足不同的客户服务质量要求。O研究人员还将开发视频流同步和合并方案，以协同的方式将网络范围的视频流与代理辅助的视频传输相结合。研究人员计划在拟议的研究中使用的方法将是理论分析、模拟研究和实验研究相结合。研究的最终结果将包括实时视频流的新概念、高效和可扩展的机制和算法、模拟和实验研究的报告以及视频代理服务器原型系统。

Many networked applications such as electronic commerce, distance learning, digital libraries, multimedia teleconferencing and on-line entertainment involve real-time delivery of stored video to a large number of clients across a heterogeneous internetwork. For continuous playback at the client, quality-of-service (QoS) must be provided in an end-to-end manner. The design of efficient large-scale video delivery systems is complicated by such challenging issues as the highly bursty variable-bit-rate (VBR) compressed video, the underlying heterogeneous networking environments, disparate client capabilities, and diverse client QoS requirements. Solutions addressing these issues must be both efficient and scalable.The main objective of the proposed research is to design and develop efficient and scalable streaming techniques for large-scale delivery of stored VBR video. The project relies on a novel distribution-tree based video delivery system architecture which employs a group of proxy servers strategically placed inside an internetwork to perform network-wide video streaming and distributed media control. Based on this system architecture, the researchers plan to develop and investigate efficient and scalable real-time video streaming mechanisms that can effectively accommodate the heterogeneity in both the underlying network environments as well as the client end systems, while providing flexible QoS supports to a large client population. In particular, o The researchers will develop network-wide video smoothing techniques which make use of the in-memory buffers at proxy servers and clients to reduce rate variability in video transmission and accommodate network bandwidth constraints. o The researchers will develop proxy video staging techniques which exploit the disk storage space of proxy servers at network edges to achieve significant bandwidth reduction in the backbone wide-area network (WAN). o The researchers will develop adaptive rate and error control mechanisms such as selective frame discard and selective retransmission which leverage application-specific information to meet diverse client QoS requirements. o The researchers will also develop video stream synchronization and merging schemes which integrate network-wide video streaming with proxy-assisted video delivery in a synergistic manner.The methodology the researcherse plan to use in the proposed research will be a combination of theoretical analysis, simulation study and experimental investigation. The end results produced from the proposed research will include new concepts, efficient and scalable mechanisms and algorithms for real-time video streaming, reports of simulation and experimental studies, and a prototyping video proxy server system.