[c-nsp] tx-ring-limit tuning

[Adam Maloney]

> Has anyone experienced these sort of symptoms before that might be able to 
> shed some additional light on the subject?

Yep.

I had a 10M VC going 7206VXR <-> 7206VXR.  When the link approached 75%
capacity, latency went through the roof.

We worked with cisco and the ATM provider extensively on this.  Changing
tx-ring-limit was their second step (first step was, of course, to blame
the ATM provider).  After that, they had us move to 12.1T for per-VC WFQ.
WFQ didn't solve it, and the image was unstable.  Cisco more or less ended
up telling us that this was the way ATM "worked", and we were SOL.  From
that point on we just maintained a larger PVC than we actually needed.