[c-nsp] 7604/sup32 (minor correction)

Ian Cox icox at cisco.com
Tue Jan 8 12:50:56 EST 2008


At 06:32 PM 1/8/2008 +0100, Peter Rathlev wrote:
>On Tue, 2008-01-08 at 07:45 -0800, Mark Kent wrote:
> > So, I'm looking at the cisco web pages and I see the 7600 is pushed
> >  big-time as a service provider edge device, and yet I see that the
> >  sup32-3b has a 300Mhz processor, and so it is not much faster than an
> >  NPE-300 (262Mhz).
> >
> > I stopped taking full routes on NPE-300 equipment a couple of years
> >  ago, moving to an npe-g1.   So, what's the scoop with the
> >  7600/sup32-3b?    It seems like a step back to me, other than the 8
> >  built-in gigE ports.
>
>The 7600 does hardware forwarding of most traffic (depends), and thus
>does not use the processor for that. It's the PFC/DFC that does the
>forwarding. That's an advantage. :-)

It forwards in hardware if the FIB will fit into the hardware FIB 
table. The full internet table will not fit into a 
PFC2/PFC3a/PFC3b/FFC3a/DFC3b/PFC3C/DFC3C anymore since the hardware 
table, even after changing from the default will only get you 239k 
routes. For full routes one needs to use a SUP7203bXL / RSP720-3CXL, 
with XL DFCs. If you don't use XLs and then you wind up with FIB 
exception and packets forwarded in software.


Ian

> > I'm looking at an application where the box would push a total of about
> >  1Gbs over two gigE upstreams.  It would have two gigE internal
> >  neighbors, each with full bgp routes... so four full tables. I'm
> >  concerned about the issue of traceroutes looking bad as they pass
> >  through the box (which confuses EndUsers), due to the cpu load from
> >  the bgp scanner.
>
>On Tue, 2008-01-08 at 12:14 -0500, Jared Mauch wrote:
> > Simple answer: Sup32 is not viable for full bgp tables anymore.  Enjoy
> > it as a l2 device that can't do 10GE!
>
>With 512MB memory, it can't take a full table? The world is moving fast
>I guess. :-)
>
>Anyway, you could adjust walker intervals etc. and maybe cope with the
>tardy processor, but it's probably not a good idea.
>
>Regards,
>Peter
>
>
>_______________________________________________
>cisco-nsp mailing list  cisco-nsp at puck.nether.net
>https://puck.nether.net/mailman/listinfo/cisco-nsp
>archive at http://puck.nether.net/pipermail/cisco-nsp/



More information about the cisco-nsp mailing list