[c-nsp] DS3 mux issues

[Pete Templin]

Justin Shore wrote:

> Ok, that's what I was thinking.  I don't know if we'll need to do any 
> fractional T1s at this point.  The majority of our offering will be LRE 
> and we'll be using bonded T1s where we can offer LRE yet so I'd expect 
> multiple bonded T1s per customer and no fractional T1s.
> 
> If we start to strain the 7206 I imagine it will be replaced with an ASR 
> or we'll pick up another 7206.  If the product offering is well-received 
> and we move into newer, more dense areas we may start with a 7600 right 
> out of the gate.  Not now though.

I have one 7206 (non-VXR, NPE-225) handling ~15 fractional T1s and ~18 
full T1s.  There are three MLPPP bundles within those ~18 full T1s.  CPU 
is generally 20-30%.  I have another 7206/NPE-225 handling ~7 fractional 
T1s and ~22 full T1s.  CPU is generally 15%, so I think MLPPP is a 
performance hog.  Our future growth will be 7206VXR/NPE-G1, and we're 
going to keep an eye on CPU so we know if/when we need to switch to the 
PA-MC-2T3-EC (which supposedly offloads the MLPPP functions to the PA).

Our growth plan isn't firm (but it doesn't need to be).  It's likely 
going to be ChOC12/T1-ISE in our 12000s, since we have one in any 
relevant market for DS3 and OCx ("high-speed WAN distribution router"). 
  The ASR1004 is also on the radar.  ASR1002 has horrible density if 
you're going to do card-diverse uplinks (there goes 2/3 of the SPA 
slots).  ASR1006, assuming 2xGE and 10x4xChT3, ends up with more than a 
GE of traffic across the T1s (likely never an issue, but...).

I'd be leery of using an Ethernet switch (7600) for high-density T1 
aggregation, but that's just me.

pt