[Beowulf] Re: The Walmart Compute Node?

[David Mathog]

Chris Dagdigian <dag at sonsorol.org> wrote:
> Subject: Re: [Beowulf] The Walmart Compute Node?

> It is dangerous to project *your* particular use cases and workflows  
> upon the community at large.
> 
> Most of the clusters I end up building or working on (academic,  
> government and corporate sites) are intended to support periodic  
> spikes in computing demands.

Exactly.  It's not just academic.  This mode of parallel operation comes
up any time the user needs the result to proceed with a project, and
either their time has value or time is critical.  So cutting a job
down from 2 hours to 6 minutes even occasionally can be justified
economically even if the cluster sits idle the rest of the time.  BUT,
while it is sitting idle you'd naturally like it to burn as little power
as possible.  The one thing that the C7 processors offer that the main
line AMD and Intel processors don't, at least last time I looked, was
the ability to run really, really, really slowly, and so drop standby
power on the processor to something negligible.  Clearly it would be
better if the AMD and Intel processors kept their capabilities and added 
equivalent low power modes.


Vincent Diepeveen <diep at xs4all.nl> wrote:

> Would it be interesting to open a discussion how to reduce sound of  
> cheapo clusters as well?

Been there - this is a VERY hard nut to crack.  The problem
isn't "cheapo", the problem is peak power and getting rid of it.  Sound
insulation is very closely related to heat insulation.  The net result
is that it's relatively easy to cool a cluster, or to muffle its sound,
but fiendishly difficult to do both at the same time.  The cheapest
way to "quiet" a cluster is to move it to a place far away from the
people who would be disturbed by its noise.  In our case, that was a
separate machine room (which already had a chilled water line for
cooling).  

Knowing what I do know, if I absolutely had to put a cluster in a
location close to users (as in, inside an office space) I'd build a big
sound insulated box with a large AC coiling coil inside it and big
(large diameter) fans to drive the air around in a defined path.  Then 
run the freon lines out to a condenser outside the building.  Notice
that this is still, essentially, moving the cooling away from the users.
 It's relatively easy to contain sound inside a box if the only holes in
it are the small ones for the power cable,the AC coolant, and a drip
line for the coil condensation.  Note that this is essentially the
approach Sun took with it's cluster in a shipping container products,
although I believe those are water cooled. Run a building air cooling
duct into a box containing the cluster and you'll have lots of fun
keeping the noise out of the building's ventilation system and from
spreading to nearby offices.  

Regards,

David Mathog
mathog at caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech