[Beowulf] Servers Too Hot? Intel Recommends a Luxurious Oil Bath

Eugen Leitl eugen at leitl.org
Wed Sep 5 07:53:13 PDT 2012


On Wed, Sep 05, 2012 at 09:54:55AM -0400, Prentice Bisbal wrote:
> On 09/05/2012 06:22 AM, Eugen Leitl wrote:
> > On Tue, Sep 04, 2012 at 02:54:46PM -0400, Ellis H. Wilson III wrote:
> >
> >> I know we've been taking things to the uber-scale level with this
> >> conversation, but does anyone have suggestions for small (homebrew
> >> Beowulf) clusters?  I've considered oil before, but for all the
> > A major advantage of these forthcoming ARM server systems is that
> > they are air-coolable, and in fact even convection-aircoolable,
> > if you add a suitable funnel on top of the rack.
> Existing systems are already air-cooled by convection.  Anything that is 
> being cooled by the movement of air is convection* When we use fans in 
> our servers, it's forced convection. Did you mean natural convection, 
> where there the air density difference (caused by the temperature 
> difference) causes the movement of air, without any fans?

Yes.
 
> Where does the funnel come in? To reduce drag that would hinder the 
> natural convection?

Sorry, chimney, not funnel. To generate draft, via a column of hot air. 
The same mechanism as in an oven.

Speaking of which, do you know how a rocket mass heater (supposedly) works? It has
no chimney but only horizontal exhaust. The only way how that makes sense is something 
like a stationary scramjet.
 
> Sorry - I spent waaaay too much time studying transport phenomena in 
> college.
> 
> * For the science nerds: Technically, convection is heat transfer by 
> flow of any fluid.

Possible, and Wikipedia seems to confirm that usage, but in the informal 
use convection cooling means passive transport as opposed by forced 
transport, e.g. via fans.

Will try to be more precise in future.


More information about the Beowulf mailing list