From: Jonas Lindberg (jonaslin@kth.se)
Date: Tue Oct 26 2004 - 17:35:10 PDT
Review of D. D. Clark, S. Shenker, and L. Zhang's "Supporting Real-Time
Applications in an Integrated Services Packet Network: Architecture and
Mechanism"
By: Jonas Lindberg
In this paper the authors present an ISPN architecture that provides both
datagram traffic and two types of real-time services: the more traditional
guaranteed service and the adoptable predicted service. A substantial part
of this paper is used for explaining why we need more than one type of
real-time service and how the predictable service can be implemented.
Simulation results presented in the paper shows that the predicted service
is useful and can be better than WFQ for applications that adapt to
bandwidth changes.
The paper is interesting and well structured. Concepts are clearly explained
and the authors do a good job pointing out what is important. The idea of
adaptable real-time applications and the design of the predicted service are
good. I think the most important contributions this paper makes are showing
how a router can provide the three services mentioned above and the
strengths of this implementation.
The presented architecture is very interesting, but a major issue is how to
deploy a thing like this. I think saying that the want for more
differentiated pricing will make ISP's deploy it is to make it a little bit
too easy. More discussion on this would have been interesting, but maybe
that is another paper?
This archive was generated by hypermail 2.1.6 : Wed Oct 27 2004 - 01:35:40 PDT