【作者】 常潘；

【导师】 张卫；

【作者基本信息】 华东师范大学 ， 计算机应用技术， 2003， 硕士

【摘要】 随着Internet的发展，愈来愈多的应用要求数据能可靠地从一个发送方传向多个接收方。在这种情况下，可靠组播比经典的传输层的单播协议更适合这种一对多的通讯。通常Internet上的应用涉及到成千上万个用户，因此如何使协议根据网络以及用户的数量具有自适应性，也就是说如何使可靠组播协议具有可扩展性一直是人们致力解决的问题。特别是那些对时间敏感的应用要求这些可扩展的可靠组播协议能够提供较短的差错恢复延时与较高的带宽利用率。然而，在通常的应用中，差错恢复延时与带宽的利用率是一对相互制约的因素，前期的可靠组播协议虽然在可扩展方面作了一些努力，但仍不能从根本上解决反馈风暴与重复接收的问题，使带宽的利用率相当低下。 本文从组播的基本概念出发，讲述了各种组播路由协议的原理以及优缺点，给出基于源路由树与共享树的组播路由协议的总结，并详细分析了目前可扩展性可靠组播所存在的问题以及前期的一些方案对这方面的改进和局限性。然后，本文采用轻量级组播服务的思想，改进了一个在文件分发中广泛采用的可靠组播协议TMTP，并命名为HRMS(Hierarchical Recovery using MBTrace and Subcasting)。它采用TMTP中基于层次的概念，通过对路由器网络层一些简单的扩展，使差错恢复聚集树与组播传送树保持一致性，同时采用一种高效的差错恢复方式——局部组播，有效地解决了反馈风暴与重复接收的问题。文中最后通过模拟的方法比较了几个改进的可扩展的可靠组播协议，表明可扩展性可靠组播协议HRMS具有较短的差错恢复延时与较高的带宽利用率。更多还原

【Abstract】 With the development of current Internet, more and more applications require data to be delivered reliably from one sender to many receivers. Under this circumstance, reliable multicast protocols are better suited for this one-to-many communication than conventional transport layer unicast protocols. The applications in Internet often involve hundreds to tens of thousands of recivers, so how can we make the reliable multicast protocols adapt to network and membership changes, that is to say how to make the reliable multicast protocols more scalable is the core problem that many researchers try to solve. Especially in the time-sensitive applications, they require the scalable reliable multicast can provide both low recovery latency and high bandwith utilization. However, the recovery latency and bandwith utilization are a great tradeoff. Although the previous researches on reliable multicast protocols have made some efforts on the scalability, yet it can not solve the implosion and exposure problem on the basis, which reduces the bandwith utilization.This thesis began with the basic concept of multicast, described the principles, merits together with flaws of most IP multicast routing protocols and gave the summary of source specific routing protocols and shared tree routing protocols. Then the author analyzed the problem on current scalable reliable multicast protocols and the improvement as well as limitations of previous works in great detail. Afterwards.the author adoped the light weight multicast service ideal and improved the Tree Based Multicast Transport Protocol which was used widely in data dissemination.In this thesis,the new protocol was named HRMS (Hierarchical Recovery using MBTrace and Subcasting) that based on the hierarchy which was used in TMTP. HRMS added a little extension to the network layer of the multicast routers, which made the logical recovery fusion tree consistent with the multicast transmission tree, at the same time, it employed an efficient recovery multicasting mode named subcasting, all these eliminated the problems concerned with implosion and exposure. At the end of the thesis, the author gave the simulation results of several reliable multicast protocols that did some improvement on scalability.We can conclude that HRMS makes a better tradeoff between the packet delivery latency and the traffic load.更多还原