h a l f b a k e r yYou gonna finish that?
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
We are reaching the limits of air for use in cooling computer systems. Air doesn't hold a lot of heat and using it to cool servers typically creates a lot of hot spots due to the complexities of fluid dynamics inside a compuer case. A large amount of space in a typical server case is taken up by fans
and empty space to allow for airflow in order to keep components within their operating temperature ranges. If heat densities continue their trend of increasing, air cooling cannot be sufficient.
A mineral oil based non-immersion cooling would allow for cheaper cooling as well as more processing power in a given space. Space in a data center is very expensive, so the costs required to implement this type of cooling may be very easily recovered. Data centers typically chill air to 20C or below in order to maintain component temperatures between 40C and 60C. Liquid cooling is much more efficient. Mineral oil is the coolant of choice because unlike water it is not conductive. Mineral oil, the primary ingredient in baby oil, is fairly safe for human exposure in case a leak should occur.
This system would require changes to both the data center and the servers themselves.
Server cases would have to be totally redesigned. Coolant would be piped to heat sink blocks in the server, rather than immersing the whole server in stagnant oil as is typical. The server would have 3 connections. One for the coolant input, one for the loop output, and a drain connector which would prevent small leaks from creating a mess as well as allow condensation to drip out. Boards would be mounted vertically and the heat blocks would be designed to drip condensation into the drain pan. Coolant inflow and outflow would be monitored so leaks could be corrected. Moving parts such as disks would be bays with heat blocks which are otherwise sealed, to prevent the oil from getting into them in case of a leak.
The data center would also require modifications for this system. Coolant loops would be run throughout the data center. Drains would be required to collect the condensation. It is likely that the coolant would only have to be chilled to 40C, but cooled further in order to keep the air temperature comfortable to humans working around the machinery.
Redundancy could easily be obtained by extra chillers, valves and loops in the piping.
While there are many potential downsides to this cooling setup, it seems to be fairly minimal compared to other alternatives to air cooling.
Current water blocks
http://www.xoxide.com/viense.html These blocks are currently used in liquid cooling and should easily adapt to other liquids [ftzdomino, Dec 31 2005]
Dry water
http://science.slas...id=04/04/14/1621235 It's called Sapphire, and it was developed for putting out fires. [Canuck, Jan 01 2006]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Destination URL.
E.g., https://www.coffee.com/
Description (displayed with the short name and URL.)
|
|
The only problem is the viscosity of mineral oil changes greatly depending on temperature. You'll have a hard time pumping cold fluid through the water blocks. |
|
|
Depends who you put it on, I guess. |
|
|
Last time honey was on me it was hot. |
|
|
Choose your words carefully [bris] or people may get ideas :) |
|
|
How about a liquid for cooling that won't get stuff wet if it leaks? See link. |
|
|
Do you mean [bristolz] covered in warm custard? Or custard as a coolling medium? |
|
|
Might as well go with an immersion system with all the redesign work this would require. More efficient. Non-conductive mineral oil is widely used as insulation for high voltage transformers and circuit breakers, so the low voltages in computer equipement would present no problems there. + |
|
|
Couldn't the cage itself be lined with cooling tubes, much like a refrigerator? That would eleminate the need for coolant i/o ports on the server. Obviously, the cage would have to be redesigned to have solid walls and insulation, but overall I think it would probably be more economical for the NOC to cool only the space that is needed. |
|
|
Overall, I'd say this idea is perfectly half-baked. Bun. |
|
|
Wouldn't ammonia (whaich has a higher specific heat capacity than water or oil) be a better coolant?
Actually something with high thermal conductivity AND heat capacity is what you want - some of the liquid metal alloys are liquid at room temperature and have fantastic thermal conductivities - an alloy of Gallium and Indium for example |
|
| |