FW: [c-nsp] Cisco Gigabit Ethernet SwitchModule (CGESM)fortheHPBladeSystem

[lists at hojmark.org]

> It now seems that our guys testing the CGESMs with the most
> current feature set are seeing that they don't seem to permit
> manual config of 100/full (and subsequently negotiate down to 
> half in many cases) and they're seeing nasty buffer overruns.

Of cause they hardcoded both end? It's a very common mistake to
hardcode only the switches, and yes that causes lots of problems.

Anway, in my experience, manually setting the speed and duplex
(at least for servers) is no longer needed, nor is it a good
idea.

While I can't think of any good reason to do so, if you really
want to limit the blades to 100, do 'speed auto 100' instead of
fixed config.

> I assume there's no HP people on this list, but why on earth
> did they chose a 2900 to repackage?

The 2970 is really a 3750, just without the routing and stacking.
So, in all practical sense, the blade switches are based on the
best low-cost platform that Cisco has. (The same as the 2960 and
3560).

Of cause they could have rebuilt a 4500 or a 6500 to house blade
servers or something like that, but that's hardly feasible when
you want something that must plug into an existing chassis.

-A