Based Multicast Overlay Network Key Technology Of Video Transmission

The rapid development of Internet and the increasing matureness of digital video technology provide the application of the large-scale real-time audio and video media streams with a rare opportunity for development. However, the Internet is the global Internet that is based on packet-switch technology and TCP/IP protocols , and its IP layer does its utmost to deliver data by means of Best-Effort and unicast communication. For the application of real-time streaming media with large data capacity and delay sensitive characteristic, these limitations of Internet technology will give rise to many problems such as the difficulty to ensure QoS and low scalability of the server.At present, the use of multicast technology can effectively solve the bottleneck problems of video server. It can effectively reduce the server bandwidth consumption and increase the scalability of server through the realization of one-to-many transmission of video data. However, due to the limitation of that most Internet backbone routers and other core equipments do not support multicast technology and the difficulties in charging, management, and security and so on of ISP, the current multicast technologies can only be achieved in a small area without the conditions for large-scale implementation. Multicast overlay network is a good substitute technology for network layer multicast, which achieves a similar function to network layer multicast through setting up a virtual network on application layer. Multicast overlay network makes the application of large-scale deployment multicast on internet become a reality.Performance Optimization of multicast overlay network should be considered from two sides: optimization of multicast overlay tree and optimization of underlay topology-aware. Designing overlay topology matched with underlay physical topology is very effective to improve the performance of multicast overlay network, but the accurate detection of the underlay topology will result in large amount of detection expenses and the huge demand of infrastructure nodes to predict the node coordinates, moreover, it is difficult to be widely implemented in the real network.In addition, QoS assurance is not considered at the beginning of the design of Internet. At present, the strategies of QoS assurance developed by IETF are put forward aiming at to data flow end-to-end transmission link. It is very difficult to achieve network-wide QoS assurance on Internet, which is as such a complex global network.In response to above problems, this thesis makes a in-depth study on the key problems existing in video transmission network from the three parts of video transmission network, namely, user access networks, transmission networks and video server. The author is committed to resolving the problems of low cost performance, poor scalability and difficulty in QoS assurance and so on existing in present video transmission systems.First of all, as for that the scheduling program of standard Patching is not suitable for hot video-on-demand, the author designs a adaptive segment-based Patching (ASP) algorithm combined with the video segmentation commonly used in server. The greatest feature of APS is to achieve a maximal sharing of regular multicast stream and Patching stream. That means, the follow-up patching request for the same video can share the preceding several or all of the adjacent Patching channel data during transmission. Simulation and analysis results show that, ASP can greatly improve the server's scalability, especially for popular video-on-demand of high rate requests. This scheme can get better performance than Standard Patching and Double Patching schemes.Secondly, a Patching flow scheduling strategy running on multicast overlay network, Range Patching, on the basis of the study model of multicast overlay network. First of all, construct a overlay topology with a two-stage hierarchical structure, and then build a single-source multicast tree on the mesh overlay topology. The use of cache resources of the overlay node to store data of regular multicast stream and patching stream can provide follow-up requests with Patching service, so as to effectively reduce the server bandwidth consumption. Simulation results show that Range Patching can greatly reduce the bandwidth consumption when the segment data is small, especially when the segment data is smaller than the average height of multicast tree, the server only need to allocate one channel for each video request. When the request rate is low, the Range Patching scheme can get better scalability than Standard Patching, Double Patching and Batched Patching schemes. When the request rate is high, the performance of Range Patching program is dependent on the relationship between average height of multicast tree and number of segment. The higher the multicast tree's average height is, the better the performance of this scheme is. In addition, Range Patching compared with Batched Patching and Pyramid Broadcasting schemes with the same technology, realizes the "real" video-on-demand service of users zero initial waiting delay.Thirdly, as for performance optimization and detection of overlay network, the author proposes an underlay topology detection scheme based on node integration. The scheme takes into account the characteristics of video transmission system and the structure of Internet, and implements node integration during topology detection, so as to effectively reduce the frequency of detection. The simulation results show that the scheme greatly increases the efficiency of detection while ensuring the accuracy of detection.Finally, as for the problem that the current BE network can not provide effective QoS assurance, the author designs and realizes a local access network bandwidth resource management system LRM, which can effectively solve the complex problems of IntServ / RSVP, DiffSrv and so on. It realizes guaranteed forwarding service through the effective management of available bandwidth of local access network and the real-time monitoring to the user video stream requests. Test results show that LRM can effectively meet the needs of local bandwidth resource management requirements, meanwhile, it can provide video stream transmission service with better service quality assurance than the traditional BE network.