[Beowulf] Outdoor location?

[Robert G. Brown]

On Tue, 22 Aug 2006, Eric McCaughrin wrote:

> In Northern Calif., we are blessed with natural,
> year-round air conditioning. Ambient temps very rarely
> go over 85F, or lower than 40F. With HVAC being a
> major (biggest?) cost, I am wondering if anyone
> considered an outdoor location?
>
> In my backyard, I have an unused 70A 240v line, and a
> large 30" high deck. Maybe this is a totally stupid
> idea, but what would be the feasibility of mounting
> nodes underneath the 2x6 planks of the deck? There
> must be some way to keep the nodes dry without
> compromising air circulation? And running the nodes
> continuously should build up enough heat to prevent
> dew from forming.

   a) Dust

   b) Bugs (and I don't mean in your code)

   c) Theft

   d) Water/Humidity

   e) Temperature variation

   f) Too high a high temperature

   g) Circulation

The outdoors is generally dirtier than the indoors, so you have to
beware of dust, pollen, and other particulate matter that can build up
and kill cooling fans etc more outdoors than in.

The outdoors is also home to many animals that aren't terribly good for
computers.  A few daddy longlegs spiders sucked into your cooling fan
won't improve it, nor would web fragments.  Snakes sometimes like nice
warm places as well.

Putting some thousands of dollars outside in an unlockable (or easily
openable) environment is an open invitation to have them stolen.  Hope
you trust your neighbors...

Doesn't it rain all winter in Northern California?  Heat may help keep
the systems dry, but it also can warm moist air to where mildew grows
happily.  Most system specs call for a fairly dry environment.
Obviously any sort of contact with liquid water is totally out of the
question, as well, so one hopes that there is no possibility of a
drainage flow or drip down from the deck in the wettest, nastiest,
windiest weather you can imagine ever getting, or you'll be trying to
unplug electronic equipment and bring it indoors in the middle of it.

A 45F temperature swing is far from ideal -- even a diurnal swing of
20-30F isn't ideal.  Parts expand and contract, including e.g. disk
platters.  These days disks seem a bit more robust than they used to be,
but in the old days if you formatted a cold disk and used it after it
got warm, the disk could actually expand enough to overwrite track
boundaries and trash the disk.  I'd still worry about this with a large
seasonal and diurnal swing, and would also worry about the effects on
certain other mechanical parts.

85F is "cool" compared to ambient outside in NC this time of year, where
we can easily reach 100F or rarely exceed it, but it is WAY hot for a
computer.  Even 75F is a bit too hot.

The effects of heat are mitigated by good circulation.  At 75F in still
air, the heat can build up and warm air can easily recirculate in a
moderately close space, raising the EFFECTIVE ambient temperature of air
getting pulled in to 85, 90, even 100F.  Baaaad.  Good circulation can
ensure that the heat is REMOVED from the systems without ever being
recirculated, so at least they are cooled by fresh ambient air.

Clusters are happiest in 65F (or cooler) clean, dry, moving air (with a
good stable airflow) maintained at fairly constant temperature, in a
secured environment.  Deviate from this in any way at some increase in
probability of moderate failure or total catastrophe (where having all
your systems stolen or actively rained on would be in the latter
category).

If it is so cool outside, then AC should be relatively cheap.  I'd
suggest putting the cluster in your basement (if you have a basement)
and putting a small window unit somewhere were it can dump that heat
into the nice cool outdoor ambient.  If it is as cool or cooler outdoors
than your target indoor temperature, the AC will work amazingly well.

    rgb

>
>
>
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
>

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu