Reliable multicast for Core-Based Multicast Routing

可靠组播是在Internet上部署组播业务的一个重要而又具有挑战性的问题。在本提案中，研究人员建议为基于核心的组播路由设计、实现和评估一种可靠的组播方案，其目标是避免NAK内爆和重复应答，减少恢复延迟，实现恢复隔离，并提供对动态拓扑/成员变化的适应性。现有的可靠组播方案主要是基于源的组播树。它们要么不能直接部署在基于核心的多播树中，要么在直接部署时可能呈现次优性能。主要的困难在于，由于缺乏每个源的信息，基于核心的树上的树上路由器不知道如何将重传请求定向到源或某些指定的应答器路由器。研究人员建议设计一种简单而有效的方案，使基于核心树的路由器能够正确地引导重传请求，从而接近模拟拐点方法中实现的最佳恢复场景。具体来说，研究人员考虑了以下研究问题：(1)为了解决识别上游/下游接口（相对于一个源）的问题，研究人员将(i)定义一个新的path_info IP选项，该IP选项携带路径信息，并在连接中有一些数据包携带该IP选项。每个接收具有此IP选项的数据包的接收器存储路径信息以供将来使用；（ii）在重传请求中定义一个新的TP_info IP选项，以便路由器知道如何指导重传请求。(2)为了在成员关系和拓扑变化的情况下进行本地恢复，研究人员将仔细设计一种机制，使用在线错误统计收集、软状态缓存和贝叶斯分析来选择指定的主机（称为应答器），直接重传请求将转发到这些主机。路由器现在跟踪每个接口的错误统计数据，并使用贝叶斯分析选择最可靠的链路作为应答链路。(3)如[levine:98]所示，纯基于nak的方案需要在源处有无限的缓冲区才能正确操作。研究人员将设计一种延迟的ACK方法，以便源和应答者都可以清除所有组成员接收到的数据包。(4)为了使所提出的方案在存在不具有可靠多播能力的路由器的情况下工作，研究人员将设计一种机制来检测云，并适当地引导重传路由器并通过它们回复消息，而不会失去恢复隔离。提出的研究是两个协同组件的组合：在分析框架中开发所有组件机制，并通过软件系统构建和实验对其进行验证。我们将利用门控组播源代码（由门控联盟，Merit Network, Inc.开发）作为软件平台，并对所提出的可靠组播方案进行原型设计（部分在基于核心的组播路由守护进程中，部分作为内核提供的路由功能）。在俄亥俄超级计算中心（OSC）和OSU医学中心的帮助下，研究人员将使用一个数据库/服务器，以从MR和CT获得的大量患者图像为中心，作为代表性的多播应用程序，以说明原型软件的关键功能、可靠性、可扩展性和自适应特性。

Reliable multicast is an important and challenging problem which is yet to be solved in order to deploy multicast services on the Internet. In this proposal, the researchers propose to design, implement, and evaluate a reliable multicast scheme for core-based multicast routing, with the objectives of avoiding NAK implosion and duplicate replies, reducing recovery latency, achieving recovery isolation, and providing adaptivity to dynamic topology/membership changes.Most exsiting reliable multicast schemes focus on source-based multicast trees. They either cannot be directly deployed in core-based multicast trees or may render sub-optimal performance when directly deployed. The major difficulty lies in that an on-tree router on a core-based tree does not know, due to lack of the per-source information, how to direct a retransmission request toward the source or some designated replier router. The researchers propose to design a simple, yet effective scheme that facilitates routers on a core-based tree to properly direct retransmission requests so as to closely emulate the optimal recovery scenario achieved in the turning point approach. Specifically, the researchers consider the followingresearch issues:(1) To solve the problem of identifying upstream/downstream interfaces (with respect to a source), The researchers will (i) define a new path_info IP Option that carries the path information and have some of the data packets in a connection carry this IP Option. Each receiver that receives a data packet with this IP Option stores the path information for future use; and (ii) define a new TP_info IP Option to be carried in the retransmission requests so that a router knows how to direct retransmission requests.(2) To conduct local recovery in the case of membership and topology changes,the researchers will carefully devise a mechanism, with the use of on-line error statistics collection, soft state caching, and Bayesian analysis, for selecting designated hosts (called repliers) to which direct retransmission requests will be forwarded. A router now keeps track of error statistics over each of its interfaces and selects, with the use ofBayesian analysis, the most reliable links as replier links.(3) As indicated in [levine:98], a pure NAK-based scheme requires an infinite buffer at the source for correct operation. The researchers will devise a delayed ACK approach so that both sources and repliers can purge data packets that have been received by all the group members.(4) For the proposed scheme to work in the existence of routers that do not have the reliable multicast capability, the researchers will devise a mechanism to detect clouds and to properly direct retransmission routers and reply messages through them, without loss of recovery isolation.The proposed research is a combination of two synergistic components: development of all the component mechanisms in an analytic framework and their validation with software system building and experiments. We will pursue this research by taking advantage of the GateD Multicast source code (developed by the GateD Consortium, Merit Network, Inc.) as a software platform, and prototyping the proposed reliable multicast scheme (partially in the core-based multicast routing daemon and partially as the routing functions provided in the kernel). With the help of the Ohio Supercomputing Center (OSC) and OSU Medical Center, the researchers will use asdatabase/server centered on large amounts of patient images acquired from MR and CT as a representative multicast application to illustrate the key functional, reliable, scalable, and adaptive characteristics of the prototype software.