Paper Review:
IP Multicast Channels: EXPRESS Support for Large-scale Single-source Applications

Reviewer: Jie Zhou

Problem

Current IP multicast realizations are strained in supporting large scale application, because they have problems in billing, accounting, sender restrication, address space, and scalability.

Contribution

This paper proposes an extension of conventional IP multicast model. The new model, named EXPRESS multicast channels, works well for large-scale multicast applications, and provides accounting surport.

Main Ideas

  • EXPRESS can be provided as a simple modification to IP multicast service model using a small portion of the class D address space.
  • The ECMP protocol that EXPRESS uses requires no change to the host operating system, because it is implemented on top of UDP and TCP.
  • Large scale multi-source applications that are almost single-source can be built using EXPRESS channels and the relaying approach supported by a middleware layer.
  • EXPRESS can provide scalability at low cost of memory and CPU.

    Critique

    This paper provides a single-source service model to solve the problems in IP multicast realizations. Though the idea is not novel, the authors give clear explanation of this model, as well as detailed analysis of its advantages and cost. I rate the paper as 3 (modest contribution).

    The basic assumption under this approach is that the majority of multicast applications are single-source or almost single source applications. In case that multi-source multicast applications prevail, the model has to build up many channels, which will incur a lot of overhead and may cause the whole network collapse.

    Lession

    In a multi-layer architecture like Internet, new protocol could be built on higher layer so as to avoid changing existent machenisms and reduce the risk associated with completely new protocol machenisms.