Data Center Temperature

For equipment efficiency, not much. For equipment longevity, cooler is better (unless you're going from 'cooling' to 'refrigeration', in which case things can get worse in real cold too).

Are you talking about 95F return air temperature or exhaust air temperature from the server?

BTW, I would re-post this in forum403. Computer engineers don't normally design HVAC plants. If you do start a new thread, click the 'Red Flag' button and ask the site admins to delete this one to avoid fragmented discussion.


----------------------------------

If we learn from our mistakes I'm getting a great education!

Yeah, thought I'd get more attention from the Comp Eng group on the air temperature here.
Good to know that a leaving temp of 95 from the server/room would not be a major problem. I know that at a certain point elec equipment runs less efficiently due to temp, and it only gets worse as it goes up.

75ºF supply seems absurdly HOT for a server room, I think.

Just make sure your system or the servers never breaks down, because whoever has to spend any time in there will want to come after you.



TTFN

FAQ731-376

That's because people of yore thought they needed to keep their machines reliable for the long term, since a 10º change can mean up to a factor of 2 degradation in MTBF. Since server life cycles are probably only a few years, since they'll be absurdly slow compared to newer servers.

Nonetheless, I think 75ºF is too hot to do much in, other than just walking around. Do any real physical exertion like ripping open a server to swap boards or whatever, and you can build up a goodly amount of sweat.

TTFN

FAQ731-376

Sheesh, I prefer a room to be in the high 70's for my comfort (77-78 is perfect for me). When it starts to hit the mid-70's and below it becomes uncomfortable, by the time the high-60's hit my fingers start to cramp.

I would disagree with nwhome's comment about the exhaust temp not mattering. If I were to pump in 40 degree air and it still comes out 100, I need more efficientcooling. The intake temp is only partially relevant (relevant only in the sense that it should be somewhat normal, such as 65 degrees and not -30), its the exhaust temp that matters, as that is what tells you (indirectly) the temp of the equipment itself. If I pump in 70 degree air, but it's coming out at only 80, I would say my equipment is staying reasonably cool... if it comes out at 100, it is really cranking out the heat and is most likely having difficulty transferring much more to the atmosphere giving its current heatsinking capability.


Dan - Owner

talking about rack inlet/outlet temps..

I saw this guys presentation at a trade show once and it was pretty informative. He has this method called the rack cooling index.

I have always made it a basic rule of thumb that if you feel comfortable in a datacentre, then its too warm. I keep mine at 18 deg C ambiant temp.

Serviced by two ceiling, omni directional A/C units, with forced air venting to the cabinets.

Hmmm this is one that always generates discussion - it is being accepted that data halls are going to get warmer as power densities are getting greater. trying to maintain a set temperature when these densities are getting higher requires more involved cooling methods to be employed.

Im starting to see integral chillers being used and also a move towards water based cooling in the halls themselves. Chilled water being fed into the rack to provide local cooling. Now that is scary, even though they say they are using better designed pipes and couplings and all that I still feel very wary that water is being introduced to a data room. 20kw-40kw in a cabinet is a huge heat load to try and dump to air without letting the temperature creep up.

I've been looking at HP's Dynamic Smart cooling.

Any thoughts on using this to control each rack to what you need or equipment requirements?