Data Center Air Distribution

Brad White_9

There is almost too much to cover, but here are some basic principles:

1)If possible, used a raised floor for supply. Floor diffusers and openings are patterned to the equipment to be cooled. Strive for at least 18" depth of not 24". 12" depth is fraught with balancing problems plus all the cabling.

2) If a raised floor is not used for supply, they make good return plenums. Units sit on the floor and draw from it. Supply can be ducted of course. Key is distribution. These rooms tend to be changed around alot and you need cooling focussed here and there and it will change.

3) If you use perforated supplies as you are describing, they are good for laminar flow (straight down if from the ceiling) but you have to keep the face velocities down, around 100 FPM max. Use "more smaller" rather than "fewer larger" diffusers. When you use that type of distribution, the return patter becomes a factor. Velocities are so low that the return airstream can effect it. Chances are you will be returning into the face of your unit.

4) Redundancy. These facilities tend to be "mission critical" so a backup scenario is key. Two units at 100% ducted into a common manifold is one way. Using three (3) units each at 50% is another. With one out of commission you still have 100% capacity (N+1). This also allows nice part-load control, with staged units. Just make sure that each unit sees the same duct resistance and has a good back-draft damper (check valve) for when a unit may be off-line.

5) Flexible ducts are a good way to get spot cooling. You have to know the equipment loads, not just what but where. Some require much cooler environments locally. Some may even require ducting.

6) Back to the upflow floor supply issue: Some racks have top muffin fans discharging upwards. They need the supply to come in low. If you supply from the ceiling, the air has to fall through the space then be drawn in, then back up. The air will have been warmed somewhat by then.

7) Humidity- reheat and dehumidification. Be clear on the needs. Most of these computer AC units cool well on the sensible side but do not dehumidify well. Coil fin spacing is almost sensible-only. You need a source of cooler dryer air (ventilation air, sub-cooled) if there is a people/latent load.

Just a few thoughts...

Unknown

Computer Room Air Distribtution

I was hoping I could get some advice on a Data Center and machine room I am currently working on. In the Data Center we will have a single 16 ft long cold aisle which will require up to 30 tons of air conditioning, or 9000 cfm. The aisle is 4 ft wide and air distribution will be overhead. I am having a difficult time determining what type of air distribtion devices to use. The ceiling is at 11 ft. I was thinking of using a row of perforated registers, with cores removed (ie a perforated return used as a supply). This will dump the air vertically, and minimize horizontal throw. I cannot determine the performance of a perforated supply as such, but feel that it will have a shorter throw than any oter device. What do you think? How would you handle this situation? Any advice would be greatly appreciated.

Thanks,

Dave

Weezbo

couple ideas...

one is take a look in a control room at an aeroport.....

look at a large computer room in a high rise with many computers...there are many things to control room environments...

i have a pdf somewhere that might lend itself to your current endeavour...i will go see if i can find it in my computer...no promises as i am not :Killer" at computers and tend not to have a highly organized eh structured retrieval system *~/:)

for now you can look at Control Resources Incorporated....
Adaptive cooling solutions..

i have looked into 40 places so far and 10 websites...so now i will go look in another section of my computer.. the PDF i am looking for is cold walls with a hydronic solution that almost completely negates the use of huge A.C and air handling equipment...

not having a lucky streak here as it were

i recall reading something at HVAC.comm not too long ago...perhaps it has something in the post that will help me remember where the file may be..i looked through my emails.:(

Bitpipe.com


this is familiar for some reason... i havent found the PDF though and i have been burning some daylight at it

well,...Today's another day and i have to go to work..i shook the computer out clean. i cannot find the pdf. the entire thrust of the project was instead of having to build heated air removal corridors in some nameless massive high rise, the mans solution was hydronics . he used enough of it, in fact, to basically have removed the a.c. tonnage equipment... there by saving almost a half a bazillion dollars to cool the place. unlike these systems the electrical costs previously were very high in other words the people doing maint on the electrical components were working in a comfortable environment so it not only reduced operational costs it also reduced maint costs and time to work on the equipment . i apologise for not being able to find where i read the white paper on the building in question. a salient point of interest is that it(the serverdata room) comprised the entire two floors of the high rise building ...

clammy

liebert

Dave i hope i spelled it right ,from my past experience all the main frame room i have worked on the company i work for usually would install a liebert split system which is just about all downflow units ,most computer room i have worked in have a raised floor they would install perforated 24 x24 computer room floor tiles and where the main frame sat they would remove floor tiles or cut them allowing air to circulate through the unit ,liebert units would pull return from the top of the unit ,a;so they would always have moisture sensor tied into the alarm contacts on the unit ,liebert are also avaiable with dx coils and back up chilled water coils ,most all of the main frame rooms i have worked in had at least 1 or 2 back up cooling systems ,in some small server rooms i have also installed larger mitsisusbi cieling mount unit which liebert also makes very nice the benefit of them is no duct work so no cooling lose through ducts and less floor space taken up ,in the recent past i have seen less liebert units used mainly because they are expensive i would probalby try to opt for a couple of either recessed cieling mount (depending on cieling grid )or suspended units mitisubushi has units up to 3 1/2 tons but plese remenber that when using them there is no cooling loses due to the fact that there is no duct work a 30 ton load sounds a little large to myself but you have to see how many btu's these unit kick out beside your room and people loads check out liebert they make the best original computer room cooling systems most larger companies have them in there computer rooms ,sony ,bear sterns ,ibm and probaly nasa peace and good luck clammy

Jerry Scharf_4

power density

The power density of modern computer rooms are insane. It's easy to have 8-10KW power/heat in a single rack that is 22" wide, 26" deep and 7' tall. Rack these up 6 wide and on both sides of an isle, and the localized heat load is off the charts. Normal chilled air distribution systems just don't work at this power density. So get some very detailed information about the power density before you start on any given approach.

Humidity control becomes a major issue as well when driving this much cooling.

jerry