CAREER: Bandwidth Allocation Techniques for Video-on-Demand Systems

随着网络和多媒体技术的不断发展，新的基于视频的应用，如数字图书馆和视频点播服务将变得越来越普遍，这需要压缩存储的视频流的高效网络传输。然而，对于稳定质量的视频流，网络资源的有效分配受到几个因素的影响。首先，压缩的数字视频数据比文本数据有更大的带宽需求，因此，对网络的带宽需求要大得多。其次，压缩的数字视频数据在多个时间尺度上可以表现出显著的带宽需求突发性。特别是，存储的视频内容通常会比“实时”视频会议类型的视频流复杂得多，从而导致更大的长期突发性。最后，由于数字视频的连续性，可能需要对网络抖动和延迟进行严格的保证（特别是视频直播传输）。为了有效地为压缩视频流的传输分配网络资源，引入了带宽平滑技术来利用视频流可能可用的先验信息。目前提出的带宽平滑技术没有考虑到带宽平滑算法与网络传输机制的相互作用。提出的研究将在两个关键领域解决带宽平滑技术与当前网络机制的集成问题：1。ATM网络的带宽平滑- ATM网络是一种流行的网络技术，因为它提供了一些最低速率保证以及不同类别的服务。在本研究中，我们将研究带宽平滑技术与ATM的三种类型：恒定比特率（CBR）、可变比特率（VBR）和可用比特率（ABR）服务之间的相互作用。在第一部分中，我们将研究带宽平滑算法到CBR和VBR业务的有效映射。在此子任务的第二部分，我们将研究带宽平滑技术到ABR服务类的映射。特别是，ABR服务类提供了最小小区速率保证，并且在网络负载相对较低的情况下，提供了一个可用的小区速率，允许应用程序发送超过所保证的最小小区速率。我们将研究将视频流塑造成ABR最小小区速率的技术，同时利用可用的小区速率在可能的情况下增强视频流。最佳努力网络的带宽分配技术——目前许多在最佳努力网络（如Internet）上传输视频的方法是通过监控网络反馈和改变视频帧速率或视频帧质量来适应可用资源。对于存储的视频流，这些算法通常不利用存储视频中可用的先验信息，导致交付给用户的视频帧率比必要的更可变。在本研究中，我们将研究在尽力而为的网络中有效地传输存储视频的技术，这些技术利用先验信息平滑传输给用户的视频帧率。本项目中的解决方案将使用两种方法进行评估。首先，将通过使用捕获和压缩视频数据的模拟研究来评估和改进解决方案。在该项目的后期阶段，我们将实现一个简单的流媒体视频传输系统，以证明所提出的解决方案的有效性。我们将利用(1)互联网、(2)连接俄亥俄州六个机构的俄亥俄州计算和通信ATM研究网络（OCARNet）以及(3)vBNS——我们可以通过俄亥俄州超级计算中心访问的超高速骨干网来评估拟议的解决方案。这个项目的贡献将是回答以下问题：在跨网络传输存储的视频流时，什么是利用先验信息的最佳方法，它可能提供不同的网络带宽保证？通过对网络资源的智能管理、客户端缓冲区以及存储视频流中可用的先验信息，本项目开发的解决方案将使我们能够向用户提供最高质量的视频。我们期望这项工作能够成为未来存储视频应用发展的催化剂，如数字图书馆、远程学习和视频点播。

As networking and multimedia technologies continue to advance new video-based applications such as digital libraries and video-on-demand services will become increasingly more common, requiring the efficient network transmission of compressed stored video streams. For constant-quality video streams, however, the efficient allocation of network resources is complicated by several factors. First, compressed digital video data has much larger bandwidth requirements than textual data, and hence, place much larger bandwidth requirements on the network. Second, compressed digital video data can exhibit significant burstiness in bandwidth requirements at multiple time scales. In particular, stored video content will typically be much more complex than "live" videoconference type video streams, resulting in much greater long-term burstiness. Finally, due to the continuous nature of digital video, stringent guarantees on network jitter and delay may be required (especially for live-video transmission).To efficiently allocate network resources for the delivery of compressed video streams, bandwidth-smoothing techniques have been introduced to take advantage of a priori information that may be available for a video stream. The bandwidth smoothing techniques that have been presented thus far have not considered the interactions of bandwidth smoothing algorithms with network delivery mechanism. The proposed research will address the integration of bandwidth smoothing techniques into current networking mechanisms in two key areas:1. Bandwidth Smoothing over ATM - ATM networks are a popular networking technology because it provides some minimum rate guarantees as well as different classes of service. In this study, we will investigate the interactions between bandwidth smoothing techniques and three of ATM's classes: constant-bit-rate (CBR), variable-bit-rate (VBR), and the available-bit-rate (ABR) services. In the first part, we will investigate the efficient mapping of bandwidth smoothing algorithms into the CBR and VBR services. In the second part of this subtask, we will investigate the mapping of bandwidth smoothing techniques into the ABR service class. In particular, the ABR service class provides a minimum cell rate guarantee and, in times of relatively low network load, an available cell rate which allows the applications to send more than their minimum cell rate that has been guaranteed. We will investigate techniques for shaping video streams into the ABR minimum cell rate while taking advantage of the available cell rate to enhance the video stream when possible.2. Bandwidth Allocation Techniques for Best-Effort Networks - Many of the current approaches for delivering video across best-effort networks such as the Internet do so by monitoring network feedback and altering the video frame rate or video frame quality to fit within the available resources. For stored video streams, these algorithms typically do not take advantage of the a priori information available from stored video, resulting in video frame rates delivered to the user that are more variable than necessary. In this study, we will investigate techniques for the efficient streaming of stored video across best-effort networks that take advantage of the a priori information to smooth the video frame rate delivered to the user.The solutions in this project will be evaluated using two methods. First, the solutions will be evaluated and refined through simulation studies using captured and compressed video data. In the later stages of this project, we will implement a simple streamed video delivery system that will demonstrate the effectiveness of the proposed solutions. We will evaluate the proposed solutions using (1) the Internet, (2) the Ohio Computing and Communication ATM Research Network (OCARNet) which connects six institutions within the state of Ohio, and (3) vBNS - the very high speed backbone network that we have access to through the Ohio Supercomputing Center.The contribution of this project will be to answer the following question:What is the best way to take advantage of a priori information in the delivery of stored video streams across networks, which may provide varying guarantees of network bandwidth?Through the intelligent management of network resources, a client-side buffer, and the a priori information avail-able from stored video streams, the solutions developed in this project will allow us to deliver the highest video quality to the user. We expect that this work can serve as a catalyst for the development of future stored-video applications such as digital libraries, distance learning, and video-on-demand.