[Beowulf] Stable power supplies
daniel at labtie.mmt.upc.es
Mon May 31 12:25:50 PDT 2004
We've recently solved our really annoying problem of random hangups,
there were two memory modules that started failing after a *long* run
period. Moreover there were some badly base polished heatsinks...
anyway, all seems running fine at the moment.
There is another remaining factor as someone suggested, the power
supply, not only the amount of watts is important but also how stable is
it running at full load. The voltage lines ( as reported by "sensors"
program ) seemed to diverge a bit more than the expected on few nodes (
on the limits as marked by "sensors" ).
Power supplies *as is* don't give an accurate information, is there any
good source of information about power supplies and testing ? Any
reccomendations about manufacturers based on experience would alse be
Last point... wondering if results reported by sensors library ( and
sensor hardware too ) are as accurate as they seem .
Daniel Fernandez <daniel at labtie.mmt.upc.es>
Heat And Mass Transfer Center - CTTC
c/ Colom nº11
UPC Campus Industrials Terrassa , Edifici TR4
More information about the Beowulf