Slashdot Mirror


Keeping a Data Center Cool on the Cheap

jedimaud writes "You've heard of bubble wrap, and the boy in the bubble -- now, here's a datacenter in a bubble. I work for a government agency that, like most, is trying to cut back some costs, and one of those costs is a REAL datacenter. So, we decided to wrap the whole thing in plastic (including two 1.5 ton ACs). The room hovers about 83 degrees, however, the racks in the bubble (ok, more like a termite tent) stay about 10 degree cooler. Here's some pics to check it out."

20 of 319 comments (clear)

  1. Hmmm by Captain+Splendid · · Score: 4, Funny

    Well, if your site is hosted by that Data Center, it just got a lot hotter in that bubble!

    --
    Linux, you magnificent bastard, I read the fucking manual!
  2. Warning: Suffocation risk by dangitman · · Score: 5, Funny

    Do not place datacenter over head. Keep out of reach of children.

    --
    ... and then they built the supercollider.
    1. Re:Warning: Suffocation risk by Mister+Impressive · · Score: 5, Funny

      Do not eat datacener.

      --
      Let the commencement BEGINULATE!
    2. Re:Warning: Suffocation risk by Knara · · Score: 4, Funny

      Do not taunt the happy datacenter

  3. Note from the metrology freak by Guppy06 · · Score: 5, Informative

    "(including two 1.5 ton ACs)"

    "Ton" here refers to a "ton of cooling," a measure of power. It was originally intended to mean "the power required to melt one ton of ice in 24 hours." Since that varies based on a bunch of conditions, it was pegged at 12,000 Btu/h.

    When they changed the definition of "calorie" to mean 4.1868 J, converting Celsius to Fahrenheit and grams to pounds gives us a conversion factor of 1 ton of cooling being exactly 3 516.852 842 066 7 W.

    In other words, each a/c unit is about 5.25 kW of cooling each, or 10.5 kW total.

    Oh, and 83 degrees Fahrenheit is about 301 kelvin and a ten-degree Fahrenheit difference is a difference of 6 kelvin.

    (According to my old HVAC prof, there's been little to no progress in "metricizing" the industry in the US. Having used both systems in his course, I'd say I prefer US units, if only because the unit descriptions on things like insulative properties make more sense when the units for thickness and area don't naturally cancel each other out.)

    (And it could be worse. Most home a/c units are labelled on the box as putting out x number of Btu, suggesting they're disposable.)

    (Well, they probably are...)

    1. Re:Note from the metrology freak by Solder+Fumes · · Score: 5, Funny

      That was entirely tangential and ultimately useless information that adds no value to this discussion. ...which means that it was perfect for Slashdot! Keep up the good work!

    2. Re:Note from the metrology freak by Anonymous Coward · · Score: 5, Funny

      (including two 1.5 ton ACs)

      Hey, I may be a fat geek, but I'm not THAT fat!

    3. Re:Note from the metrology freak by Guppy06 · · Score: 4, Informative
      " but last time I checked, a nine degree Fahrenheit difference was a difference of five degrees Celsius,"

      ... which means 10 is 5 5/9, rounding up to 6. Guess I didn't use the word "about" enough times for you.

      But hey, if you really wanna get anal, do it right.

      "assuming that 273 K = 32 degF,"

      32 degrees Fahrenheit is 273.15 kelvin.

      "The temperatures chosen were freezing and boiling of water at 1 atm pressure, in case anyone didn't know."

      At 101 325 Pa, water boils at 373.124 kelvin. Celsius /= centigrade. The freezing point is also flakey, but not within three decimal places at least. This is why all modern temperature scales are pegged to the triple point (273.16 K)instead of trying to correlate two data points that move around with respect to each other.

      "The correct Fahrenheit boiling point of water (at 1 atm) is 212 degF.)"

      Try 211.953 degrees Fahrenheit.

      Oh, and the first letter of SI units (like kelvin) is never capitalized.

  4. Real Data Center by aaarrrgggh · · Score: 4, Informative

    First off, a "real" data center needs a little more than 1.5 Tons of cooling; that will barely cool a single rack in a real data center.

    Insulation is always a nice idea, be the fact of the matter is that to reject the heat from the space you need to provide a means of heat trasnfer. Generally, that requires a temperature differential between your heat source and the outside. If it is cooler outside than in the space, not much is required. If it is warmer, you will have to take advantage of thermodynamics and use a compressorized cycle. This can be more or less efficient, depending on the difference in inside and outside temperatures.

    (A typical data center operates with a 95-110F outside design temperature, and attempts to deliver 48F chilled water to the CRAC units (Computer Room Air Conditioning). This forces about 50% of the energy consumed by the computers to be used (again) to cool the equipment.)

    Call an engineer when everything melts down...

  5. Re:Build the datacenter in alaska by Solder+Fumes · · Score: 4, Insightful

    Heck, why not? Run some massive fiber along the big pipeline. Only disadvantage is that you're out on the edge of the Internet, you're not safely embedded in the middle with nice redundant capabilities.

  6. It's misleading by SamMichaels · · Score: 4, Insightful

    1.5 ton sounds expensive, big and awe-inspiring. It's not. Most people have cheap 1 ton a/c units in their living room walls (12,000 btu). My 12,000 (1 ton) unit is barely able to cool 3 computers. Good luck with a datacenter.

    1. Re:It's misleading by drmerope · · Score: 4, Informative

      No offense, but maybe you missed the point?

      The article was entitled "Keeping a Data Center Cool _on the Cheap_" (emphasis added)

      The suggestion being that they were able to significantly cut back on the size and power consumption of the cooling plant by using plastic wrap.

  7. this doesn't work... by Anonymous Coward · · Score: 5, Funny

    keeping a datacenter cool? Thats nothing..

    try moving to india to try and get your job back from dell only to get dysintry and heat stroke, lose your wallet and end up working in low grade indian miget porno to get enough money to buy a can of coke, only to get typhode.

    Bubbles...pfff...

    Ps. I'm writing this from Iran, send help.

  8. Re:Wonder if... by shawb · · Score: 4, Insightful

    I think in this case the AC only has to work on pumping the excess heat from the servers out, and not cooling the whole room. Doesn't matter how warm the room gets outside of the bubbles.

    --
    I'll never make that mistake again, reading the experts' opinions. - Feynman
  9. Re:Needs more duct tape... by Leroy_Brown242 · · Score: 4, Funny

    and more cowbell.

  10. Heh, great minds think alike! by pla · · Score: 4, Interesting

    Having recently faced a similar problem (though on a much smaller scale), we came up with almost the same solution.

    As one suggestion, though, cardboard (in 4x8ft sheets) proved a lot easier to work with than plastic sheets. For starters, the plastic requires attachment at the ceiling, and will eventually come loose under its own weight; cardboard, with a single fold in the sheet, will stand upright and support its own weight for years, assuming not too high of a humidity level. For another, cardboard won't flap around and potentially block air intakes nearly so easily as plastic will.

    Believe it or not, though, what we found the most effective way to make use of barely adequate AC - Don't treat the room as a closed system. You've basically used the plastic sheets to build giant chimneys - Now take advantage of that fact, and along with a high volume fan above each rack, just exhaust the air at the top outside rather than recycle it back into the room... Think of it this way... You spec your cooling to work to perhaps 110F ambient, right? At the top of a full rack, with 50-60F going in the bottom, you probably have 120-130F going out the top. Does it take more work to cool 130F, or 110F, back to 50F? Not to mention, your normal ambient shouldn't come anywhere near 110F...

  11. Re:Build the datacenter in alaska by atarione · · Score: 4, Funny

    the real problem with a data center in alaska is that the fucking polar bears will keep drinking all the cherry coke.

    --
    actually I am happy to see you, however that is in fact a banana in my pocket.
  12. Note from the Entomology freak by morcheeba · · Score: 4, Funny

    "(more like a termite tent)"

    "Most termite species are tropical or subtropical, but a few live in temperate regions." I'd posit that even fewer live in a properly cooled data center. So, on the surface (no pun intended), this doesn't seem to be a good comparison.

    But reading further into the Wikipedia article: "Termites have biting mouthparts and are soft-bodied, of moderate to small size. They live in dark nests and tunnels, except when the winged alates emerge to leave their parent colony." When comparing termites to geeks, they both seem to have biting mouth parts and the geeks are definitely soft-bodied. And of smaller size. And, like the termites, true geeks live in their parent's basement.

    "Termites cannot themselves digest the wood that they consume." Few geeks can live on chewed-up pencils. So again, another similarity.

    Lastly, Termites construct extremely large and elaborate mounds to house their colonies. ... The internal structure of these mounds can be quite complex, with ventilation chimneys for active temperature control" Need I say more?!

  13. Don't try this in YOUR datacenter. by deacon · · Score: 5, Funny
    Hey Kids!

    Ever wonder why computers are made of metal? Why the plastic pieces are made of plastics that meet flammability performance standards?

    Tiny Tim raises his hand

    Yes, Tiny Tim?

    Please, Mr Deacon Sir, so that if some source of ignition is present, the computer does not turn into a fireball spewing deadly poisionous smoke, Sir!

    Very Good, Tiny Tim. And what happens if someone hangs up huge sheets of generic, flammable plastic in an area with lots of potential ignition sources?

    Please, Mr Deacon Sir, sooner or later it catches fire, and people die. If these boneheads are lucky, Sir, someone from the Building Facilities or the Building Inspector will see this website and make them take it down, Sir.

    Very good, Tiny Tim. The rest of the class is to read up on Flammability.