[c-nsp] Data Center cooling

[Zoe O'Connell]

Michael K. Smith - Adhost wrote:
> We are in Seattle and use an air-exchanger system that relies on outside
> air as much as possible, and then blends in chilled water as necessary
> up to 100% chilled.  It's fairly common here because of the nature of
> our climate, and the psychrometric scale
> (http://en.wikipedia.org/wiki/Psychrometrics) is favorable for us.
>
> We've also looked at increasing our data center temps from 68F/20C to
> closer to 78F/25.56C (hi Gert), but our marketing folks have been the
> most resistant because of the prevailing expectation that colder is
> better.  There is some good research and testing being done by
> Microsoft, Intel and Google in this arena, but I don't think enough has
> been published yet to give that calming feeling to the marketing folks.
> I would imagine, however, that we will see increasing data center
> temperatures more and more in the coming years.

This also depends on how well you're circulating the air within your
data centre - having air at 25°C is fine as long as all that air
actually reaches the things it needs to cool. If it's been mixed in with
enough hot air by the time it's got to the top of the rack at the far
end of each row however, you're going to run into trouble.

Closer to the original topic, I do recall seeing a TV programme some
time in the last few years that mentioned cooling the computer room at
some Antarctic science base and they did still have to use compressors
etc as it was easier than trying to make the outside air suitable,
although I forget the details. (I suppose, at least, you could dump the
warm air into the rest of the base but I seem to recall the computers
were in a separate hut/building)