[c-nsp] Thanks (Re: Sup720 vs RSP720 - Difference?)

Gert Doering gert at greenie.muc.de
Fri Jun 19 14:23:14 EDT 2009


Hi,

(I'm copying back my response to c-nsp, because it ended up longer than
intended, and it might be useful to have in the archives)

On Fri, Jun 19, 2009 at 11:10:43AM -0700, Rick Ernst wrote:
> Thanks for the tremendous help you've given on the Sup/RSP question.
> I've been wading through white-papers, spec-sheets, Google, CCO, etc.
> trying to get my brain wrapped around what's going on, and none of it has
> been as useful as the information you've provided.
> 
> As a note, I'm going from all software routing (7206VXR/G1, 7507/RSP16) to
> 7600 series, so my brain is not yet calibrated for proper understanding
> and knowing which questions to ask. :)

There's a tremendous wealth of information in the archives of cisco-nsp,
as "us others" have had the same startup confusion as well.

There are a few important things to keep in mind:

 - if a "software router" is unhappy with something, it will get "somewhat
   slower" because it's going to be executed in a slower software forwarding
   path - but in the end, it's all "software".

 - if a (Cisco) "hardware router" is unhappy with a combination of features 
   you enable, the performance will go down *drastically*, because the
   hardware is extremely fast and the CPU on these boxes is fairly weak
   (the Sup720 is slower than a NPE-G1).  So check the set of desired
   features first - some are just not very suitable for fast-but-dumb
   devices.  NAT is one of the border cases, reflexive ACLs are tricky,
   and one of the worst thing is "tunnels with fragmentation".

   Most of this is documented, though.

 - the 6500/7600 series is "a big switch with extra brains".  This means
   that it will be less flexible in some cases than a "real router" - the
   most notable thing is the global VLAN space.  This means that if you
   have "dot1q vlan 2" on one interface, and "dot1q vlan 2" on another
   interface, it will be the *same* vlan 2.  On a "router", it's two 
   different dot1q subinterfaces, while on the switch, it's "two trunk 
   ports that carry the same VLAN (2)".

   The positive side of this is that you can play much nicer tricks with
   ether-channel aggregation than with "routers" - like the GSR that still
   can't do all possible features on an ether-channel (for the longest time,
   no IPv6 support at all on ether-channels...).

   This is really the most important thing to keep in mind: the 
   architecture is much closer to a switch than to a "classic" router, 
   and this has upsides and downsides.


 - there are the SIP and ES cards that plug into the 6500/7600, and 
   effectively bring their own brains - read: different bugs, different
   features, and different behaviour regarding VLAN space and such.

 - if something you want is not shipping today, don't believe any of the
   promises they are going to make.  Especially regarding combinations of
   line cards, chassis types, and supervisor boards - customers have been
   badly burnt by Cisco internal fights here.  Make them sign that this
   is going to work or else they will be taking back the boxes.


gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 304 bytes
Desc: not available
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20090619/51570910/attachment.bin>


More information about the cisco-nsp mailing list