Sorry for the silence, but I think you can infer from it that no one has
tried this.
From discussions we've had with people who have considered this sort of
thing in the past, here's what I can remember:
You generally have no control over the number of interfaces and switch
topologies of these things: if each blade has, say, 4 NICs, those NICs
are going to be wired in a topoloy that may or may not be what you want.
You generally have very little choice in switches, and those switches
might, even though they are from vendors we support, require 'porting'
of snmpit to them. They may or may not be fully provisioned; eg. you'll
have to look closely to see if you're going to get artifacts from
under-powered switches.
You'd need to port our power control and serial console support to them,
and from what I hear, some blade servers might not give you per-blade
control: eg. you might have to reboot entire shelves, not individual
blades, or might only be able to use the console on a single blade in
the shelf at a time.
Now, these caveats may or may not be true of the specific blade platform
you're looking it, but it'd be worth looking for them to make sure they
are not present or that you can at least live with them.
Thus spake Jonathan Walsh on Tue, Dec 15, 2009 at 07:15:02PM -0500:
I was wondering if anyone has had any success in using HP ProLiant blades
as Emulab nodes? These would be an HP BladeSystem ProLiant Chassis with 8
HP ProLiant BL460c G1 blades. I guess as a more general question, what
are the implications of using a blade center for testbed nodes? I would
guess there would be some complication depending on how the network back
plane for the chassis was set up for creating vlans, etc. as well as power
control, and console connectivity.
Thanks,
-Jonathan
_______________________________________________
Testbed-admins mailing list
Testbed-admins@flux.utah.edu
http://www.flux.utah.edu/mailman/listinfo/testbed-admins