Reviewer: Charlie Wiseman
Date: 11-17-2005
How would you rate this paper, relative to others we have read? top 25%, but not top 10%
How would you rate your kowledge of the topic of this paper? novice
What problem or issue does the paper address? Why is it important?
This paper discusses (multi-match) packet classification, which is needed for NIDSs and packet accounting schemes.
What are the main contributions of the paper and why are they important?
The packet classification scheme introduced here (SSA) uses both less memory and less power than previous solutions. These two properties mean that the overall cost of a switch needing packet classification can be lowered dramatically (less build cost and less operating cost).
How significant are these contributions relative to previous work?
Not being familiar with this area, it is hard to say exactly how important SSA is. However, the authors claim that it is the first such scheme that maintains comparable speeds to similar schemes while lowering the power and memory needs of the switch. At least compared to the other schemes mentioned in the paper, this appears to be true, making SSA a significant step in the right direction.
Give detailed comments justifying your view of the paper.
The results seem quite promising. In particular, the results shown indicate that SSA meets all of their requirements nicely. For example, SSA uses almost as few TCAM entries as possible, while not increasing the overall number of TCAM accesses too much. This leads to lower memory requirements and lower power usage. In addition, updating the rule set appears to be relatively cheap, although there is the added overhead of having to re-run the SSA. As with all such hardware schemes, for it to be fully accepted there should be an actual piece of hardware built using SSA for packet classification to show how well it performs.
The comparison to software solutions seemed relatively week, although I have the (possibly false) impression that it would be very difficult for a software scheme to compete with a hardware one at reasonably high speeds. So, this doesn't seem like a big hole.
One minor annoyance is their lemma about SSA. While not really important, they aren't using generalized terms for their intersections and clauses (they use F1,F2,etc). Just change the subscripts to x,y,etc.