[c-nsp] MWAM module on 6500/7600

[Kristofer Sigurdsson]

On Tue, 2005-11-08 at 08:21 -0500, Rodney Dunn wrote:
> The MWAM was originally developed for mobile wireless stuff.
> 
> Then folks realized it was "sorta" like 6 NPE-G1's on a blade
> so they could hack it up in all kinds of different solutions
> like (mGRE termination, some broadband usage, high density
> GRE termination for large neighbor counts, etc..).
> 
> However, it's not the exact same as 6 NPE-G1's. I'd use
> 6 7301's instead.

> It's easier to manage in my opinion because the packet flow
> and troubleshooting to/from that blade can be pretty hard
> to follow. To and from a stack of 7301's is much easier to
> follow in my opinion.

Assuming static IP's for the clients, clients distributed between
the boxes in ways that make summarization impossible, the six 7301
way would probably mean distribution of host routes (in this case
up to 32,000 host routes to the network) *or* for all of the six 
routers to announce a summary route and then distribute the traffic 
among themselves (meaning that 5/6 of the traffic reaching each box is 
meant for another one, creating load) *or* for at least two of the 
7301's just being dedicated as upstream routers for the broadband 
aggregation boxes (announcing a summary route "out" and taking in host 
routes from the aggregation boxes).

This means you need eight 7301's to achieve the same as you could use 
one 7600 + MWAM.  If memory serves me right, the 7600 + MWAM combo 
costs roughly the same as three 7301's...also, the 7600 is a much more
scalable box, feature and throughput-wise...

There's always the argument about putting all your eggs in one basket,
but you could probably buy two 7600's with MWAM's for the price of that
7301 farm, getting a fully redundant setup.

Am I missing something here?

-Kristofer