基于Mesh的P2P流媒体协议研究.doc_第1页
基于Mesh的P2P流媒体协议研究.doc_第2页
基于Mesh的P2P流媒体协议研究.doc_第3页
全文预览已结束

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

基于Mesh的P2P流媒体协议研究摘要 流媒体应用是互联网和多媒体技术发展的趋势和前沿性研究课题。基于Mesh的P2P网络凭借其高可扩展性和部署成本低等特点,成为了解决流媒体应用的最重要的途径之一。论文主要针对基于Mesh的P2P流媒体的关键技术展开研究,主要包括覆盖网的构建、数据通告策略、流媒体调度以及激励机制等方面。 论文首先系统概述了P2P流媒体的相关研究工作。详细介绍了几种P2P覆盖网络中的资源定位方式和路由模型,并给出了一些代表性的协议和软件。并根据P2P流媒体中的覆盖网拓扑所划分的两种P2P流媒体模型:基于树型的流媒体和基于Mesh的流媒体进行讨论,分析它们的各自的优缺点,得出基于Mesh的流媒体相对于基于树结构的优越性,同时分别给出了此两种模型的代表性协议,并对当前基于Mesh的P2P流媒体协议存在的一些问题进行了探讨。 近乎最优的带宽利用率使得BitTorrent对流媒体传输来说是个很好的选择。通过分析BitTorrent协议与P2P流媒体协议的相似性,论文提出了一个基于BitTorrent的P2P流媒体协议。为了使得BitTorrent协议适合流媒体传输,我们对该协议进行了如下修改。(1)基于数据分片的最后播放时限,我们给予它们不同的优先级别。(2)源节点采用push-based的机制来分发数据。(3)我们引入了block-level tit-for-tat策略来取代BitTorrent中基于速率的tit-for-tat策略。实验结果表明改进后的BitTorrent协议能有效利用节点带宽,提供好的服务质量并具有很高的可扩展性和鲁棒性。 针对当前基于mesh的P2P流媒体协议中存在的不足之处,论文第四章提出了一个基于Mesh的P2P流媒体协议: FairStreaming。在FairStreaming中,我们首先提出了一种数据块通告策略,源节点向其邻居节点通告不同的可用数据信息,避免了传统的通告策略在低带宽环境下数据的严重丢失。其次我们提出了一个P2P流媒体中数据块调度的简单模型,揭示了数据块调度中,节点的下载速度与数据块大小、缓冲区大小、节点带宽等的关系。并根据该模型,我们提出了一个流媒体调度的贪心算法,根据数据块在邻居中的分布状况和播放时限赋予它们不同的优先级,并优先请求具有最高优先级的数据块,使得这些数据块总是能在最短的时间内得到。最后,我们提出了一个MIMD(Multiplicative Increase Multiplicative Decrease)激励算法,每个节点计算其与每个邻居节点上传下载的数据块数目,并根据上传与下载的数据块数目之差来决定其下一轮向该邻居节点的发送速度。该算法能有效促使节点贡献出自己的带宽,驱逐自私节点。通过模拟实验等手段,我们对该协议进行了性能评估。实验表明,FairStreaming具有很高的带宽利用率,并且即使在低带宽环境下也能提供很高的服务质量,同时能有效驱逐自私节点,保证了全局吞吐率。 关键字:P2P网络、流媒体、BitTorrent、贪心算法、激励机制Abstract Streaming application has become a hot research area as the rapid development of Internet and multimedia technology. With high scalability and low deployment cost, mesh-based P2P network has become the most important approach for streaming multimedia over Internet. This thesis hence focuses on the key technologies of mesh-based P2P streaming, including overlay construction, data revelation policy, data scheduling and incentive mechanism. In the summary of the related work in P2P streaming area, we first give an introduction of the resource location and routing model of several P2P overlay networks, together with some representative protocols and softwares. Then we discuss the advantages and disadvantages of two P2P streaming model: tree-based P2P streaming and mesh-based P2P streaming, and come with the conclusion that mesh-based P2P streaming behaves better than tree-based approach. We also introduce the representative protocols of these two models, and investigate the problems existed in mesh-based P2P streaming protocols which have been presented. The near-optimally bandwidth utilization makes BitTorrent a good candidate for content delivery. Based on the analysis of similarity between BitTorrent and P2P streaming, we present a P2P streaming protocol based on BitTorrent. To overcome the inherent characteristics in BitTorrent which are not suitable for streaming, we impose following modifications on BitTorrent protocol. First, data pieces are assigned with different priorities according to their playback deadlines. Second, the source disseminates data pieces in a push-based way. Third, the rate-based tit-for-tat policy is replaced with a block-level tit-for-tat policy to avoid low start of new peers. The simulation results indicate that the modified version of BitTorrent can effectively utilize bandwidth, provide good QoS, and has a good scalability and robustness Finally, we propose a mesh-based P2P streaming protocol: FairStreaming. In FairStreaming, we first present a strategic data revelation policy where the source reveals different available data segments to its neighbors, makes them have different partial views on the available data of source, thus the neighbors of source would request different data segments to the source, which increases the data diversity of system, and avoids the serious data loss which would happen in traditional revelation policy in the presence of low bandwidth network environment. Second, a simple model of data scheduling in P2P streaming is presented, which reveals how key factors, e.g. data segment size, buffer size, peer bandwidth, impact peers download rate in data scheduling. Based on this model, we propose a greedy data scheduling algorithm where each segment is assigned a priority based on its playback deadline and distribution among neighbors, and the segments with highest priority are always preferred to be requested, which makes these segments can be acquired as soon as possible. Finally, a MIMD (Multiplicative Increase Multiplicative Decrease) incentive mechanism is proposed to deter free-riders. For each neighbor, each peer computes the number of data segments which have been uploaded to and downloaded from the neighbor, and decides the rate of sending to the neighbor next round based on the difference of these two numbers. Through extensive simulation, we show that FairStreaming achieves a high bandwidth

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论