Review 8

From: Charles Reis (creis@cs.washington.edu)
Date: Tue Oct 26 2004 - 22:56:23 PDT

  • Next message: Seth Cooper: "Review of "Supporting Real-Time Applications in an Integrated Services Packet Network: Architecture and Mechanisms""

    Supporting Real-Time Applications in an Integrated Services
      Packet Network: Architecture and Mechanism
    Clark, Shenker, Zhang, 1992.

    The paper takes a very thorough and systematic approach to discuss the requirements for supporting real-time network applications side by side with datagram applications. The authors make several new and useful distinctions in the process, most notably between rigid (and likely intolerant) applications and adaptive (and likely tolerant) applications. This leads to different service models: guaranteed and predictive, each of which suits its respective class of applications well.

    The authors do a good job breaking the principles and mechanisms down and discussing them individually, making another distinction between isolation of flows (eg. using WFQ) and sharing of resources (eg. using FIFO+). The two service models are even proposed separately before being combined into a single scheduling algorithm.

    While the paper effectively draws out many important issues in supporting real-time applications, the actual mechanisms it proposes remain more proofs of concept than evaluated solutions. (Indeed, the authors note that evaluation of such proposals is very difficult.) It is also unlikely, even given appropriate pricing incentives, that the extra complexity needed in the central routers (and in the edges for enforcement) will be deployable. Despite this, the paper is still useful for providing a framework for discussing quality of service issues in the Internet.


  • Next message: Seth Cooper: "Review of "Supporting Real-Time Applications in an Integrated Services Packet Network: Architecture and Mechanisms""

    This archive was generated by hypermail 2.1.6 : Tue Oct 26 2004 - 22:56:18 PDT