On Fri, 2 Jun 2000, Markus Pfeiffer wrote:
> Sasi Peter wrote:
> >
> > On Thu, 1 Jun 2000, Gregory Maxwell wrote:
> >
> > > About the only think I can think of is turning off CPUs on a SMP
> > > laptop for powermanagement.. :P So unless someone from transmeta speaks
> > > up about a quad cpu laptop this just sounds like a feature looking for a
> > > need. (i.e. worthless cruft)
> >
> > No!
> >
> > Think of a fan failing on one of the CPUs, in the middle of a veri
> > important computing task. If you can umount that CPU, it will not burn and
> > crash from overheat, and your process can also finish off.
> >
> > Thinking of HA, when desinging the future of linux is a must...
>
> ehemmm,
> I think the CPU will burn even if it is umounted ...
Why? If you stop the processor, it should be generating very little heat;
enough that it won't need the fan running, at least. Just inserting "HLT"
instructions during idle moments can have a significant impact on CPU
temperature; if you can enter a hard HLT loop (i.e. no IRQs, just HLT) you
should be OK without a fan.
> and a fan failing
It happens. Moving parts eventually stop moving - long before solid-state
ones fail, usually.
> Good, if you take a very cheap one, but if one does high performance
> computing one doesn´t use crappy equipment, and even if a failure of
> expensive equipment is possible, I think the scheduler changes are too
> expensive ( of the speeds point of view ... ).
There are no scheduler changes needed. Just disable that CPU when needed;
the scheduler continues scheduling as normal (except obviously it can't
schedule anything to run on the failed CPU.)
> People at Be seem to be playing people...
I can't see the software CPU on/off switch as a major selling point,
somehow. As an HA facility, though, being able (ultimately) to hot-swap
CPUs and fans would be quite useful.
James.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:15 EST