[c-nsp] c7206VXR(NPE-G1) w/ 12.2(18)S8 - Memory failures

[Jared Mauch]

On Thu, Mar 31, 2005 at 04:35:03PM +0200, Gert Doering wrote:
> Hi,
> 
> On Thu, Mar 31, 2005 at 03:16:22PM +0100, Neil J. McRae wrote:
> > Agreed, and I think Cisco are taking steps to resolve this
> > issue but in my view its not strategic for them to do this.
> > For example we have IOX coming to a GSR near you soon (although
> > only one that was bought in the last 12 months :( ), but ask
> > about IOX for the 7609 or anoy other platform and you get 
> > "uhmm maybe, probably not", and finally, "no we are doyng 
> > anything with IOX"
> 
> And *this* is one of the worst problems.  Platform groups deciding what
> IOS trains (or "non-IOS" trains) get ported where, without listening to
> the customers.  If a 7600 is positioned as a *router* for *ISPs*, but
> then doesn't get IOS-next-generation - what's that good for?

	Something is in the works for that, you should ask your
account team about it :)

	I think the problem is that Cisco is entirely driven on
the platform side, ie: we must ship this platform or feature
or ELSE!  This leaves the IOS people stuck creating all these
forks/branches that we see littered about on the outside.

	You can see evidence of this with other software companies but
in lesser fashions, eg: New/Fixed MSIE but only for OS revs A, B and C but
not D, E.

	You are seeing the impact of these "platform" forces in other
companies as well, as slightly different platforms are released you
see a shift in the software that is required for them.

	I think the single biggest thing that Cisco needs to solve
is the lack of standardization across platforms for memory/flash.

	Some of this is the result of changing technologies (eg: linear
flash cards -> ata flash cards), simm->dimm-> sodimm -> ???

	some of these are driven by form-factor choices, but
going with a standard across the board would be ideal for the customers,
eg: Using CF cards.  They're small, they work reliably, they're
cheap and best of all, standard.

	Same goes for when manufacturing route processor boards, stick
with a standard memory type (which Cisco has done in most platforms)
but the key is to avoid artificial limits, eg: I can put a 512m dimm
in my npe-225 or npe-300, but they are unable to size past
256m.  It was forseable in the future that you would need more
memory (as always has been the case), it leaves the case where increased
effort on figuring out how to make a 512m or larger dimm be accessed
would cause customers to see longer life of their devices, causing them
to continue using them longer, instead we're seeing them end up on
ebay and sold off at much lower costs when the forwarding performance
is sufficent!

(end rant).
	- jared

-- 
Jared Mauch  | pgp key available via finger from jared at puck.nether.net
clue++;      | http://puck.nether.net/~jared/  My statements are only mine.