My advice would be to list the major points of the paper,
and then try to partition your time to make those points.
Here's my (partial) list:
1. You can construct a model of TCP performance as a function
of loss rate, RTT, etc. This model works!
2. TCP is amazingly sensitive to loss rate. It doesn't
matter what the link bandwidth is, if routers drop packets
TCP dies (Figure 4). Combined with Paxson's work that
showed that Internet drop rates were around 5%, the implication
is that you can't get good TCP performance, no matter how fast
you make the Internet.
3. You might think adding buffers at routers will help by
improving the resilience to bursts, but adding buffers
increases the RTT, which hurts.
4. RED fixes fairness issues with TCP phase effects.
<that's as far as I got, and I was only through half the paper>
You could spend 5-10 minutes on each of these major points,
and I'm sure no one would be bored. Make sure to parcel your
time so you get to all the main points. We really should have
spent 2 days on this paper.
tom