Reviewer: Ben Wun
Date: 10-6-2005
How would you rate this paper, relative to others we have read? bottom 50%
How would you rate your knowledge of the topic of this paper? novice
What problem or issue does the paper address? Why is it important?
This paper provides a new abstraction model that should be useful in analyzing the performance of router architectures that traditional models do not caputre. This paper provides theoretical calculations for the amount of resources needed to support a single buffered (SB) abstraction model in both a DSM and PSM router while achieving similar performance to CIOQ routers.
What are the main contributions of the paper and why are they important?
The main contribution of this paper is to provide a theoretical framework for thinking about SB router models and how to provision them for the best performance and for providing QoS.
How significant are these contributions relative to previous work?
The authors provide an analysis of their model and show what is needed for various types of SB routers to achieve performance comparable to the existing CIOQ model.
The authors make the claim that their model can help us analyze existing deployed routers for which the CIOQ model does not work, but they never really demonstrate this.
Give detailed comments justifying your view of the paper.
This paper does not make a convincing case for adapting the SB model.
The authors claim that many existing routers fall under the SB model, but do not tie this back to their theoretical work. They give a list of open problems, but do not mention how existing SB routers deal with them, nor do they evaluate how well their model applies to real hardware.
While it seems, in theory, that we can build a PIFO DSM router using less memory, and simpler scheduling algorithms, there are too many other problems that the authors are do not show a solution for. The increased crossbar bandwidth, global information requirements reduce the applicablity of this work. They cite related work using the constraint sets to analyse the PPS, but they presented many other ideas beyond constraint sets in this paper.