【作者】 陈金莲；

【导师】 奚建清；

【作者基本信息】 华南理工大学 ， 计算机软件与理论， 2012， 硕士

【摘要】 近年来随着互联网的高速发展和信息传输技术的快速发展以及宽带网的发展，用户对流媒体的服务质量要求也越来越高。基于P2P技术的流媒体系统成了当前研究的热点，网络电视、视频会议、远程教学和视频点播等都已经为越来越多人熟悉和使用。P2P流媒体系统对于传统的C/S架构的流媒体系统克服了服务器瓶颈问题，部署成本和复杂度大大降低，在可扩展性和容错性方面都有无法替代的优势。节点选择是P2P流媒体领域的核心技术和研究热点。由于P2P节点服务能力有限和多样性，如何进行节点选择是一个难点。本文充分考虑超级节点的作用，提出基于分区的超级节点算法。分区算法将物理位置邻近的节点划分到同一个区，可以有效地减低节点与节点间的传输延迟。然后充分考虑节点的带宽、cpu、内存等性能参数，考虑节点的负载情况，提出自适应超级节点算法，该算法根据超级节点的运行状态自动对超级节点进行调整，在超级节点负载过重时，自动更新超级节点，可以有效地提高系统的健壮性。本文深入研究服务节点选择算法，提出了基于混合P2P架构的双端协作的节点选择算法。该算法可以有效地把请求节点的数据请求分发到同一区域或者其他区域中具有请求资源的节点上，可以有效地降低资源服务器的压力，缩短节点数据传输时间，降低骨干网络压力，保证数据的连续传输，保证系统服务质量。本文还实现了基于主动拥塞机制的可用带宽测量算法和基于包对模型的瓶颈带宽测量算法，并对这两个算法进行了实验验证，然后将该算法应用到超级节点选择算法以及服务节点选择算法中。最后，基于“飞影”系统，将这些算法应用于系统的成员管理模块，并对系统的资源服务器和客户端进行测试，验证了该节点选择算法的有效性，并指出了下一步的工作。更多还原

【Abstract】 With the rapid development of Internet and information transmission technology, and thedevelopment of broadband network in recent years, the requirements of users on the quality ofservice of streaming media is increasingly higher. Streaming media system based on P2Ptechnology has become the current research hotspot, more and more people are familiar withnetwork television, video conference, distance learning, video on demand and so on, and usethese applications. As far as streaming media system with traditional C/S structure isconcerned， P2P streaming system overcome the server bottleneck problem, greatly reducesdeployment costs and complexity, has no substitute for the advantages in scalability and faulttolerance. Node selection is the core technology and research focus of field of the P2P mediastreaming. Due to the limited service capacity and diversity of P2P nodes，how to select nodeis a difficult problem.This paper fully consider super node, and propose super-node algorithm based onpartition. The partition algorithm will group the nodes of which the physical location areadjacent into the same district, which can effectively reduce the transmission delay betweennodes and nodes. Besides fully considering bandwidth, cpu, memory and other performanceparameters of nodes, and consider the load of the nodes, this paper raises an adaptivesuper-node algorithm. This algorithm adjusts the super node according to the operating statusof the super-node. When the super node is overload, the algorithm automatically updates thesuper-node, which can effectively improve the robustness of the system.This paper studies the service node selection algorithm, and proposes the double-endcooperative node selection algorithm based on the hybrid P2P architecture. The algorithm caneffectively hand out the data request of request node to the nodes that have the requestedresource on the same region or on other regions, which can effectively reduce the pressure ofthe resource server and shorten the data transfer time between nodes. This algorithm canreduce the pressure of backbone network，ensure data continuous transmission and ensure thequality of system servicerequest data requesting node distributed to the same region or other regions of therequested resource node can effectively reduce the pressure of the resource server to shorten the node data transfer time, reduce the pressure of backbone network to ensure datacontinuous transmission, to ensure the system quality of service.This paper also implements available bandwidth measurement algorithm based onself-induced congestion and the bottleneck bandwidth measurement algorithm based onpacket-pair. We have do experiments to validate these two algorithms, and applied them to thesuper-node selection algorithms and the service node selection algorithm, and note next job.更多还原