基于Bloom filter的下一代互联网可扩展组播技术研究

Design of a scalable multicast protocol is an important research issue for Next-Generation Internet, which is expected to accommodate large numbers of multimedia applications over a common IP-based transport infrastructure. The traditional IP multicast is bandwidth efficient in data delivery but not scalable in managing the multicast tree. The more recent overlay multicast establishes the data-dissemination structure at the application layer, however, induces redundant traffic at the network layer. The widely accepted new multicast protocols based on Bloom filter have good scalability; however, the practically large-scale deployment and applications of these protocols are faced with three challenges: 1) the lack of incremental deployment solution, as it is impossible to upgrade all the nodes in the network overnight to be aware of the new protocol; 2) the Bloom filter incurs the false positive, which once ocurr in some subtle cases, can result in forwarding loops and thus may cause the partial break-down of the network; 3) the lack of the research on how to seamlessly integrate the Bloom filter based protocols into multimedia applications. .To deal with these challenges, we propose to use the destination-oriented methodology in the design of Bloom filter based multicast protocols, where the key idea is that each multicast data packet carries explicit destinations information, instead of an implicit group address as in IP multicast, to facilitate the multicast data delivery; the bandwidth overhead for such explicit addressing is tackled with the Bloom filter technique. With this methodology, we plan to achieve the following goals in this project: 1) propose a tunneling-based solution for incrementally deploying the Bloom filter based multicast protocol in a network, where only a small fraction of the routers are aware of the new protocol while others are legacy routers; 2) present a loop elimination scheme based on the reverse path forwarding to completely eliminate possible forwarding loops incurred by the Bloom filter false positive; 3) design a channel zapping acceleration mechanism by exploiting the features of the new multicast protocol, in order to seamlessly integrate the protocol into the IPTV service, and reduce the channel zapping time perceived by IPTV users. The findings of the project will provide valuable references to practically deploying the Bloom filter based multicast protocols, as well as the design of scalable multicast mechanism in Next-Generation Internet.

如何设计具有可扩展性的组播机制，以支持未来海量的多媒体应用，是下一代互联网研究需要解决的重要问题。早期的网络层和应用层组播都存在可扩展性差的缺陷，目前被广泛接受的基于Bloom filter的新型组播机制虽然具有良好的可扩展性，然而其真正大规模部署和应用却面临着三方面的挑战：缺乏可行的渐进性部署方案；Bloom filter可能造成组播转发环路；缺乏此类协议如何与多媒体应用无缝结合的研究。为应对上述挑战，本项目计划实现如下目标：为基于Bloom filer的新型组播协议提出一种基于隧道技术的渐进性部署方案；设计一种基于反向路径转发的去环机制，彻底去除Bloom filter可能引起的转发环；提出一种基于新型组播协议的IPTV频道转换机制，降低频道转换时延。本项目研究结果可以为基于Bloom filter的可扩展组播协议的实际部署和应用，以及下一代互联网可扩展组播协议设计提供参考。

如何设计具有可扩展性的组播机制，以支持未来海量的多媒体应用，是下一代互联网研究需要解决的重要问题。早期的网络层和应用层组播都存在可扩展性差的缺陷，目前被广泛接受的基于Bloom filter的新型组播机制虽然具有良好的可扩展性，然而其真正大规模部署和应用却面临着三方面的挑战：缺乏可行的渐进性部署方案；Bloom filter可能造成组播转发环路；缺乏此类协议如何与多媒体应用无缝结合的研究。为应对上述挑战，本项目完成了如下研究：为基于Bloom filer的新型组播协议提出一种基于隧道技术的渐进性部署方案；设计一种基于反向路径转发的去环机制，彻底去除Bloom filter可能引起的转发环；提出一种基于新型组播协议的IPTV频道转换机制，降低频道转换时延。. 具体来说，对AOM的基本协议进行了扩展，扩展后的协议与原协议的功能具有兼容性，协议在数据包头标识出AOM数据包，使得具有AOM支持能力的网络节点可以将其识别出来。我们采用隧道技术，将AOM数据包整体作为数据载荷下发，这样就可以实现渐进式部署下的数据传输。我们在网络仿真器ns2上实现了具有渐进性部署功能的AOM协议，结果证明了我们提出方法的有效性。. 在AOM协议去环机制设计上，我们完成 AOM 协议在复杂拓扑环境下产生环路的概率分析，归纳AOM 协议由于Bloomfilter 假真现象引发环路的实际场景；实现环路消除机制，保持AOM 的可扩展性优势并证明所设计的环路消除机制有效性。. 在AOM的IPTV频道转换加速机制设计上，我们揭示了基于TSS模型的现有操作模式会破坏切换延迟的边界，并提出解决该问题的方案。我们还提出用面向目标的多播来加速切换的机制，使得订阅用户在无需任何网络控制信息交换的切换之后，可以无缝从子信道移动到主信道。我们在ns-2网络仿真器中实现了DAZA，并通过实际的网络拓扑多播了MPEG-4视频流，实验证明了我们对DAZA理论分析的正确性以及DAZA的优越性。. 本项目研究结果可以为基于Bloom filter的可扩展组播协议的实际部署和应用，以及下一代互联网可扩展组播协议设计提供参考。