Investigating the Individuality/Scalability Tradeoff in the Provision of Networked Services

网络服务是信息革命的核心。本文主要研究网络服务的两个特征：1)可伸缩性：服务以较低的单位用户成本处理大量终端用户的能力；2)个性化：服务处理由于通向终端用户的网络路径的异构性而产生的不同需求的能力。尽管高度的个性化和可伸缩性是任何联网服务的理想特征，但它们通常是相互冲突的目标，需要相互权衡。本研究旨在从网络基础设施的角度来研究这种权衡。该项目试图回答：1)在使用多播通信的服务中，是否有可能在以受控的方式牺牲可扩展性的同时增加个性化的级别？2)在使用单播通信的服务中，是否有可能在控制牺牲多少个性的同时提高可扩展性？3)是否可能不要求将服务标记为单播或多播，并将决定适当的网络支持的负担转移到“通用”接口层？为了回答这些问题，该研究试图开发网络支持机制来导航个性化/可伸缩性频谱，使多播服务具有更个性化的能力，使单播服务具有更强的可扩展性。这项研究将包括调查向多播服务添加个性化度量的具体实例，以及通过多播共享资源和在服务器上请求批处理来使单播服务更具可扩展性的具体实例。这项研究将包括视频分发、网络信息传递和多媒体点播(直播和流媒体)等服务。这项工作将考虑互联网和ATM风格的网络，重点是Internet over ATM体系结构。

Networked services are at the heart of the information revolution. This work focuses on two characteristics of a networked service: 1) scalability: the ability of the service to handle a large number of end users with low per-user cost, and 2) individuality: the ability of the service to deal with differing requirements imposed because of heterogeneity of network paths leading to the end-users. Although a high level of both individuality and scalability are desirable features of any networked service, they are normally conflicting goals that need to be traded-off with each other. This research aims to investigate this tradeoff from the networking infrastructure's viewpoint. This project seeks to answer: 1) In services using multicast communication, is it possible to add a level of individuality while sacrificing scalability in a controlled manner? 2) In services using unicast communication, is it possible to improve scalability while controlling how much individuality is sacrificed? 3) Is it possible not to require the labeling of services as unicast or multicast and move the burden of deciding the appropriate network support to a "generic" interface layer? To answer these questions the research seeks to develop network support mechanisms to navigate the individuality/scalability spectrum, giving multicast services the ability to be more individualized and unicast services the ability to be more scalable. The research will consist of investigating specific instances of adding a measure of individuality to multicast services and specific instances of making unicast services more scalable by sharing resources through multicast and requesting batching at servers. The research will include services such as video distribution, Web information delivery and multimedia-on-demand (live and streamed). The work will consider Internet as well as ATM-style networking with some emphasis on Internet over ATM architectures.