基于动态时间坐标系的P2P VoD系统关键技术研究
文献类型:学位论文
作者 | 武广柱 |
学位类别 | 博士 |
答辩日期 | 2008-06-03 |
授予单位 | 中国科学院声学研究所 |
授予地点 | 声学研究所 |
关键词 | 对等网络 视频点播 资源定位 负载均衡 调度策略 网络编码 网络仿真 |
其他题名 | Studies on Key Technologies of P2P VoD System Based on Dynamic Time Coordinate System |
学位专业 | 信号与信息处理 |
中文摘要 | 近年来,基于对等网络(Peer-to-Peer,P2P)的流媒体系统成为研究热点。P2P直播系统已经逐渐走向成熟,但P2P点播(Video-on-Demand,VoD)系统却因异步性、交互性、冷播现象等而在资源定位、数据调度等方面具有挑战。 本文对P2P VoD关键技术进行了深入研究,主要研究内容和成果如下: 提出了一种用于P2P VoD资源定位的动态时间坐标系的建立方法,并在此基础上,提出了一种基于动态时间坐标系的分层DHT拓扑组织方法DTCS-DHT,和一种基于动态时间坐标系的搜索表拓扑组织方法DTCS-ST。在动态时间坐标系下,节点播放点坐标在不进行VCR操作的一段时间内为一定值,解决了C/S或者DHT索引方式难以用于P2P VoD系统来追踪时刻变化的节点播放点的问题;DTCS-DHT具有较高的合作节点查询效率;DTCS-ST具有较好的负载均衡性质。 提出了一种负载均衡的分布式索引服务LBDIS。LBDIS通过将具有相同ID的超级节点构成群,群间再组织成Chord环的方式,使一热点索引项引起的负载能够均衡到Chord环上其后继群内的多个群节点上。 对网络编码在P2P VoD系统中的应用进行了研究,提出了一种自适应随机网络编码流媒体分发算法ARLNCStream。其创新在于节点能够自适应调节编码窗口的大小,从而提高了播放的流畅性并降低了服务器负载。 提出了一种推拉结合的启发式数据调度策略PPHDS。其创新在于缓存结构的四部分划分、片段下载速度与选择范围的控制、稀有先取同随机选取相结合的片段选择算法、片段优先权最大化的启发式算法以及父节点数据推送算法。仿真表明,PPHDS具有非常高的播放流畅性并能够大幅度提升系统的服务能力。 针对传统网络仿真平台不适合于大规模P2P内容分发类应用层仿真需要的现状,提出并搭建了一套用于大规模P2P内容分发仿真的平台FALPS。FALPS速度快、内存省、精确度高,是本文仿真试验的基础平台。 |
英文摘要 | Peer-to-Peer Video-on-Demand system becomes a hot spot of research and development nowadays. Indeed, current live P2P streaming systems have been shown to be very efficient for large-scale distribution with few server resources. However, on-demand media distribution using P2P is still a challenge. The fundamental challenge of this problem is the unpredictability of user requests: asynchrony, non-sequentiality, zipf distribution. Thus, to find, for a given peer, another peer that can provide the required streaming data is a critical issue. And finding scalable data scheduling strategy is also very difficult. This thesis gives further studies on key technologies of VoD over P2P network. The main research contributions of this thesis include: DTCS, dynamic time coordinate system for P2P VoD, is proposed. Based on it, a hierarchical DHT topology structure (DTCS-DHT) and a search table topology structure (DTCS-ST) are presented. DTCS makes a peer’s playing point coordinate constant. Thus the DHT or index server doesn’t have to be updated constantly. DTCS-DHT makes the complexity of FIND be O(1), which is O(logN) in DHT. DTCS-ST is a efficient and load balance scheme. LBDIS,load balance distributed index service, is proposed. LBDIS resolves the DHT’s load balance problems when used in P2P streaming systems. LBDIS is also very efficient. ARLNCStream, a novel algorithm of adaptive random linear network coding for media streaming, is presented. When distributing contents using random linear network coding, every node adjusts its coding window in an adaptive way and then requests coded block of an appropriate size from its partners. The solution makes nodes collaborate well with potential partners. Without knowledge of global network topology, ARLNCStream archives quiet low segment missing ratio. ARLNCStream is also robust and scalable. A push-pull data distribution algorithm, PPHDS, is proposed. Pull operations are used between neighbors. When a peer’s uplink is free, it pushes data to its children. To maximize priorities of chunks, a heuristic data schedule strategy is used. The proposed data distribution scheme helps to make sufficient utilization of end-peers’ resource, alleviates the load of source server and improves the scalability of the system. FALPS, a discrete event simulator for large scale network simulation, is designed. Backbone network model is used in the simulator. Based on this model, we present three strategies to make FALPS efficient: event combination, piece discarding and event queue size control. Results from performance experiments show that except for its extra height speed and accuracy, FALPS reduces the memory consumption significantly. |
语种 | 中文 |
公开日期 | 2011-05-07 |
页码 | 129 |
源URL | [http://159.226.59.140/handle/311008/322] ![]() |
专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
推荐引用方式 GB/T 7714 | 武广柱. 基于动态时间坐标系的P2P VoD系统关键技术研究[D]. 声学研究所. 中国科学院声学研究所. 2008. |
入库方式: OAI收割
来源:声学研究所
浏览0
下载0
收藏0
其他版本
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。