Tjernlund M-6: Finding A Better Way To Cool The Benchmarking Server Room

[phoronix]

Phoronix: Tjernlund M-6: Finding A Better Way To Cool The Benchmarking Server Room

After months of experimenting with different methods for better cooling the basement server room where 60+ systems are running Linux benchmarks on a daily basis, for less than $100 USD I've found a fan that does an amazing job keeping the temperature suitable for all of the running systems.

http://www.phoronix.com/vr.php?view=22369

[jonsmirl]

Water cool them and use it to make hot water for your house!

[Michael]

Water cool them and use it to make hot water for your house!

Unfortunately that's no where close to being economical...

[Ardje]

Cooling in a datacenter can be underrated. In fact the design of your airflow is probably the most important thing in a datacenter.
Do not ever have some cases in a room. Have a cold side, and a warm side. The best is to either isolate the cold side or the warm side.
In your case I would probably isolate the warm side, using a plastic "tent" or something like that, and suck the air from the isolated part to your livingroom, or to the outside. You will need blinders in your closet to prevent loss of pressure and getting unheated air.
Long ago there were mainframe rooms, and they were designed that the whole room has a constant temperature. But mainframes were nothing more than big closets of air with tiny computers in them.
Then the internet came, and in a mainframe room for a single mainframe, you can house > 10kA worth of pc-servers. That's when you start to cry when you just build another dataroom intended for a mainframe instead of pc-servers... And it takes another 10 years before hot and cold corridors became a thing ;-).

[Dukenukemx]

Unfortunately that's no where close to being economical...

Unless you shop on Ebay.

Linus had no problem doing it.

[Michael]

Unless you shop on Ebay.

Linus had no problem doing it.

That's five systems in the video, not around 60.... Doing so would easily be several thousand dollars.

[darkbasic]

That's five systems in the video, not around 60.... Doing so would easily be several thousand dollars.

Not counting several hundreds of work hours...

[Michael]

Not counting several hundreds of work hours...

And figuring out how to easily route tubes into the 2U cases, etc.

[geearf]

Unless you shop on Ebay.

Linus had no problem doing it.

Interesting videos, but the guy clearly says that they won't do it again, that it was a fun experiment, but not as good as AC would be.

[boxie]

That's five systems in the video, not around 60.... Doing so would easily be several thousand dollars.

And doing it at 60+ machine scale would probably warrant a more industrial solution including a water chiller, a gravity tank (up high so that the water flows down hill), a fairly serious water pump and a lot of water blocks... all to remove the same amount of heat that the current air solution does!

...Have a cold side, and a warm side. The best is to either isolate the cold side or the warm side.
In your case I would probably isolate the warm side, using a plastic "tent" or something like that, and suck the air from the isolated part to your livingroom, or to the outside. You will need blinders in your closet to prevent loss of pressure and getting unheated air...

This could be a cheap experiment to run michael... and might make things cheaper next summer... and you can try it out on 1 rack first with a few well placed temp sensors to get the data you need.