When comparing cluster offerings, seems reasonable, that the additional $85-$100 would be factored in to any system/cluster purchase, for at least power up/down and reset? This is astonishing, or is there something I'm missing in this thread? The technology mentioned isn't really earth
depends on the scale and environment. when I built my first cluster (started at ~48 nodes), I had controllable power and a serial network for bios/console. I never actually wired up the latter, and once the system stabilized, actually removed the controllable power.
for small clusters (<= few racks), neither should really be necessary, since your hang/crash rate should be very low. if you have quite fat nodes, the marginal cost becomes negligable, though, and worth the convenience. for larger clusters (say, >256 nodes), even occasional manual fiddling is dubious - may still be OK if you have lots of people.
my organization has ~2500 nodes, many remote, and ~5 admins who need to spend most of their time doing systems/net/storage/web programming, etc. so it's pretty important to avoid a lot of driving and/or labor-intensive operations... _______________________________________________ Beowulf mailing list, [email protected] To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
