节点文献
网络视频服务系统中服务机制及算法研究
Research of Service Mechanism and Algorithm for Video Service System
【作者】 吴杰伟;
【导师】 赵会群;
【作者基本信息】 北方工业大学 , 计算机应用技术, 2007, 硕士
【摘要】 近年来,随着高速计算机网络、数字压缩技术以及大容量存储技术的快速发展,基于网络的视频服务逐渐成为现实,基于网络的视频服务在娱乐、教育、广告、信息获取等各个方面都有广泛的应用。然而,这些系统中普遍存在的问题是,服务质量和服务水平受到网络带宽瓶颈和视频服务器瓶颈的限制,表现为网络拥堵,服务延迟、中断,甚至由于服务器负担过重而拒绝服务。为解决上述问题,一些视频点播(Video-on-Demand,简称VoD)系统通过提高硬件和服务器配置的方式来提高系统性能;有些VoD则通过调整系统结构,如采用了分布式或P2P体系结构把工作负载分配到多台服务器(或对等点Peer)上,以集群协同工作的方式来提高系统性能;有些系统则通过采用流调度技术,如扩展指数广播(Extended Exponential Broadcasting,简称EEB)、控制多播(Controlled Multicast,简称CM)等,在不增加硬件系统性能情况下,通过广播或信道共享等机制来突破网络带宽瓶颈的限制,从而提高系统的服务用户数量。然而,无论采用何种方式,都有其各自的优缺点和适用范围。本文在分布式体系结构的基础上采用单播、CM和EEB等多种流调度方案,实现了一个可扩展的分布式VoD系统。对实时流协议(Real-Time Streaming Protocol,简称RTSP)进行了扩展,使其能传输服务器的负载等信息。并针对该系统采用多种流调度方案的特点,提出一种对基于Linux虚拟服务器(Linux Virtual Server,简称LVS)系统的传统动态反馈负载平衡算法的改进算法,使任务负载的分配更加合理。在节目存储策略方面,本文提出了以调整节目的流调度方案来代替节目存储调整的观点,并给出了一种节目流调度方案可自动调整的实现方法。另外,本文还介绍了如何提高客户端播放器解码能力,以及非线性编辑技术在节目制作中的应用等方面的研究内容。最后,本文给出了一种系统仿真测试方法及其实现过程。测试结果表明,该系统能有效缓解传统VoD系统中服务器I/O或网络带宽的瓶颈问题,具有良好的可扩展性;改进算法比传统负载平衡算法能减少10%~30%系统并发信道占用数,提高了系统支持的用户点播数量。
【Abstract】 Network video service systems came true in resent years, along with the fast developing of the fast computing network, digital compressed technology and the large capability storage technology,. Video services based network are widely applied to recreation, education, advertisement, information acquisition and other domains, like Video-on-Demand (VOD), distant education, online trading and initiative news etc.But, there is one ubiquitous problem in these systems, that is these systems’ quality of service are restricted by the bottleneck of network bandwith and video server, so that it causes network huddle, service delay and interruption, or worse that service requests would be refused because the servers are overload.In order to solve the problem above, first, some VoD systems are improved by enhancing the systems’ hardware. Second, some are improved by adjusting the systems’ architecture, like adopting the distributed or P2P architecture, so that the work load can be distributed to multi servers or peers, which cooperate with each other and work as a cluster. Third, some VoD systems adopting new channel scheduling schemes, like Extended Exponential Broadcasting (EEB), Controlled Multicast (CM) etc., these technologies can break through the bottleneck of network and increase the amount of the users, by using broadcasting and steam sharing mechanisms. However, whatever mode they used to improve VoD system, in which there are advantages and disadvantages, and they applied to different fields.An extensible distributed system would be present in this paper, which employs some Streaming scheduling schemes including unicast, CM and EEB. We extended the Real-Time Streaming Protocol (RTSP) and used it to transfer the load informations of server, so the servers can connect with each other and form as a distributed system. Aimed at this system had different Streaming scheduling schemes, we present an improved algorithm to traditional dynamic load balancing algorithm based on Linux Virtual Server (LVS), and it made the workload distribute more reasonable. In the programs storage aspect, we present a policy that the adjusting of programs storage can be substituted by the adjusting of programs Streaming scheduling schemes, and then we present an auto-adjusting implementation of the programs channel scheduling schemes. In addition, we introduce how to enhance the decode capability of VoD client player, as well as how to use DirectShow Editing Services (DES) to make a video clip. Finally, we introduce the technology about implementation of simulation for VoD. We proceeded to test our VoD system with simulation program. The Result showed that the system could lighten the bottleneck problem in traditional systems, can reduce 10%~30% system’s concurrent streams and serve more users, with the help of improved algorithm compared with traditional algorithm, this system can satisfy large scale of VoD requests.
【Key words】 Video on Demand; Distributed; Load Balance; Streaming Schedule;
- 【网络出版投稿人】 北方工业大学 【网络出版年期】2007年 03期
- 【分类号】TP311.52
- 【被引频次】5
- 【下载频次】118