[c-nsp] sup2t -- where the deets' at?

[Nick Hilliard]

On 28/05/2010 22:27, Anton Kapela wrote:
> On 28/05/2010 21:44, Tony Varriale wrote:
>> What issues are you having on the 04 and 08s?
> 
> Nothing that cannot be avoided (mostly, that is) by planning explicitly
> the placement of ports in lacp/pagp groups vs. subint-fannout/access
> ports vs. inter-chassis links, [...]

So how can Cisco deliver this?  The root problem here is that the sup720
and 6704-10ge date from late 2002 and technology has moved on in the last
8.5 years.  Although having said that, the sup720 and 6704-10ge deserve
some level of admiration for having had such a remarkable life-span.

Unfortunately, you simply can't get a fully non-contended 4 x 10G service
on a sup720/rsp720/vss based chassis, so there's little point in Cisco
redesigning a new-and-improved 4 port blade without first having an updated
supervisor card.  And the problem with this is that if you update your
supervisor, you're forced to upgrade your entire line card portfolio,
because running mixed capability cards in the same chassis is hard without
dropping everything to the lowest common denominator.

Also, updating the sup720 to just deal with kit-facing 40G line rate would
represent lots of capex for what most people would regard as a marginal
gain in capability (i.e. 10G instead of, well, whatever you get depending
on what you're doing).  This means that the only viable option for a sup
upgrade is to go from 2x20G to n*(20+m)G, where n>2 and m is enough to cope
with bus overhead and other such things.  We expected this from the
sup1440, but that got transformed into the sup-2t.

The sup-2t is horrendously late, and really ought to have been delivered
several years ago.  Its late delivery has left the 65k/76k product line in
a position where they have been eclipsed by the competition in terms of
both raw throughput and 10G chassis density.  The 65k/76k series does lots
of things really well, just not 10G.

Nick