HyperCast - Highly Scalable Multicast Communications

Http://www.cs.virginia.edu/~jorg/research/hypercast/计算机网络中最大的挑战之一是满足对高级组播服务的需求，以支持广域网中的大规模组播应用。许多应用要求提供可扩展的组播服务，包括电子软件分发、协作工具、多方视频游戏和分布式模拟。组播应用可伸缩性的一个主要障碍是组播组成员需要彼此交换控制信息。在这个研究项目中，提出了一种新的方法来解决可伸缩多播问题，其中多播组的成员使用具有强对称属性的逻辑控制结构来组织，例如，n维超立方体。通过利用控制结构的对称性，可以有效地实现需要在多播组成员之间交换反馈信息的操作。该项目的目标是提高组播通信的可扩展性，并证明了容纳几千个成员的组大小的可行性。利用分析技术、模拟和测量实验，我们将研究构建大型多播组的不同方法的基本限制和实际权衡。将设计并实现一个在分组交换网络中维护逻辑超立方体的协议原型，称为HyperCast协议。如果成功，该项目将影响目前正在进行的设计可扩展组播协议的研究工作。该项目开发的解决方案将适用于许多需要多播通信的计算机科学和工程领域，包括并行和分布式模拟、分布式信息系统、全球虚拟数据库和元计算环境。

Web page: http://www.cs.virginia.edu/~jorg/research/hypercast/ One of the biggest challenges in computer networking is to meet the demand for advanced multicast services in support of large-scale multicast applications in wide-area networks. Numerous applications demand the availability of a scalable multicast service, including electronic software distribution, collaboration tools, multi-party video games, and distributed simulations. A major impediment for scalability of multicast applications is the need of multicast group members to exchange control information with each other. In this research project, a novel approach to the problem of scalable multicast is proposed where members of a multicast group are organized using a logical control structure with strong symmetrical properties, e.g., an n-dimensional hypercube. By exploiting the symmetry properties of the control structure, operations that require an exchange of feedback information between multicast group members can be efficiently implemented. The goal of the proposed project is to improve the scalability of multicast communications, and demonstrate that the feasibility to accommodate group sizes of several thousandmembers. Using analytical techniques, simulations, and measurement experiments, the fundamental limits and practical trade-offs of different approaches to structure large multicast groups will be investigated. A protocol prototype, called the HyperCast protocol, that maintains a logical hypercube in a packet switching network will be designed and implemented. If successful, the project will influence the currently ongoing research efforts on designing scalable multicast protocols. The solutions developed in this project will be applicable in many areas of computer science and engineering which require multicast communications, including parallel and distributed simulations, distributed information systems, global virtual databases, and meta-computing environments.