mogul, receive interrupt livelock

[Robert Morris]

Mogul's paper about receive interrupt livelock covers a lot of the
same ground that we do. His performance test methodology is very
similar to ours also. We need a sharp argument about how we're
different.

His technique does not appear to improve the peak forwarding rate;
it only prevents the forwarding rate from decreasing with overload.
Since he's turning off interrupts, I don't understand why he doesn't
get higher peak performance. Anyway, that's a possible advantage
we have over him.

Another might be that we have a more general plan for scheduling
all activities in a router. He only tries to control device handling.

We're also going to present a much more detailed analysis of why
we get the performance we do. Which has enabled us to do more tuning.

We have language level optimizations which he can't touch.

Mogul's paper is in ~rtm/papers/mogul-livelock-usenix.ps