[c-nsp] 12008 replacement, 7300?

David Barak thegameiam at yahoo.com
Fri May 13 12:22:50 EDT 2005


some comments in-line.

--- Matthew Crocker <matthew at crocker.com> wrote:
> >> Personally, I would stay with the 12000 series
> and upgrade the LCs
> >>
> >
> > The other idea is to upgrade some existing 6500s,
> turning them into  
> > 7600s. Versus the 12000 series, would you look at
> that at all?

> The SUP720 is a new beast and has limited software
> features and  
> reliability (judging from this list).

That's getting better, but it's not quite up to the
SUP2/MSFC2 point yet.

> The 6500/7600 is a switch 1st, router 2nd and has
> small packet  
> buffers. OSM cards have large buffers?
> The 12000 is a router with large packet buffers.

Watch out - each linecard has a single set of ingress
and egress buffers, which are shared between the
ports.  You can get unpleasantness if you have one
port which is overutilizing - the hot port can starve
the other ports on the card.  I have only encountered
this with E0 cards, and most frequently with the 12DS3
card, especially when the hot port was subrated.


> The 12000 does dCEF all the time,  doesn't know how
> to do anything  
> else. If you don't have enough RAM on an LC to
> handle the FIB what  
> happens?

Router = barfed.

<anecdote> I worked at $VERY_LARGE_ISP in their NOC a
few years ago, when a person mistakenly imported their
BGP routes into their IGP.  The ISPs core routers
weren't Cisco, but their edge were, and so the core
routers carried all of these changes (the withdrawal
and re-announcement) out to all of their edge routers,
and pretty much every uplink card (E0/2) barfed
spectacularly, and had to be power-cycled (thank
$DEITY for "test mbus power [slot] off")

> If you are taking big pipes and feeding little pipes
> you'll want the  
> bigger packet buffers.  high latency is better than
> dropped packets.

depends on your traffic type - there have been some
talks at NANOG about how small buffers can actually
improve TCP throughput, and how large buffers can
decrease it.  Certainly large buffers are bad for
jitter, so if you're doing circuit-em or other traffic
where Jitter is an issue, you'll want really small
buffers.

> I don't know the direction Cisco is going with
> either package,  to me  
> it looks like they are developed by two different
> engineering groups  
> competing against each other.  CSR-1 & 12ks in the
> core and 7600s on  
> the edge?? I dunno

agreed - I'm certain that there is someone who
understands the decision process which has led to so
many totally independent platforms, although it seems
like the console port is finally migrating to the
front of all of the boxes... (I'll take what I can
get...)

-David

David Barak
Need Geek Rock?  Try The Franchise: 
http://www.listentothefranchise.com

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


More information about the cisco-nsp mailing list