[c-nsp] 2811 usage scenario

[Mike Byrne]

On Fri, 12 Nov 2004, Dave Temkin wrote:

> >From my testing, I got 600Mbps bidirectional (300Mbps each way at the same
> time) of UDP traffic through the box at packet sizes around 500.  This is
> far above what the box is rated at, so I'd imagine you'd be fine.  A 3640
> would fall over well before hitting 50Mbps.
>
> -Dave

Hi Dave,

Wow, that with a 2811 router?  Any chance you did or could do the same
test with all process-switching (no ip route-cache on each interface)?

Also a show ver output would be great, I haven't seen any information on
what processor they put in the 2800 series...

Thanks.

- Mike


> On Fri, 12 Nov 2004, Mike Byrne wrote:
>
> >
> > Hi all,
> >
> > I'm looking at possibly replacing an existing 3640 with a 2811.
> >
> > We use the 3640 for FE-FE (NM-2FE-2W) and FE-ATM (NM-1A-T3) routing, and
> > while the 3800 series seems like more of an upgrade, the 2811 seems like
> > it should do what we need.
> >
> > The reason we're looking for a replacement, though, is because the CPU on
> > the 3640 gets maxxed out occasionally (bridge traffic across ATM PVCs).
> > The usage is almost all interrupt-driven, and we're maxxing out at about
> > 10mbps on the DS3 now before the CPU is at 100%.
> >
> > So the question is if the 2811 (or 2821?) can do better.  We need the
> > bridging (irb) and can't switch to rbe anytime soon.  I'd like something
> > that can do 20-25mbps on the T3 to hold us over for a while.  Is this
> > realistic with the 2800 series?
> >
> > Going by the published kpps ratings for the 3640 and the 2811 it seems
> > like a huge improvement, but without any published process switched kpps
> > ratings it's very hard to tell.
> >
> > Thanks for any input...