What Are Economical Methods to Cool Down Mineral Oil in Server Tanks?

[Anon_Miner]

I'm dating myself, but:

In the good old days, large computer installations were often installed on a false floor with a single large blower / AC pressurizing the under-floor space. Ducts were directed from that space up into the individual enclosures - the pressure difference caused the cool air to flow through the equipment.

You could build a version of that, and use a single large blower (located outside) instead of using the local fans. It's important to get the air where you need it - I'd thoroughly test with a single server before I did anything at scale. You may need to do some 'micro-ducting.'

I did something similar. I created a "tunnel" using tarp; at one end I put a commercial Evaporative cooler (15000 m3/h blowing power) and at the other end I placed the servers with their fans removed. the problem was that the servers' boards are really dense and don't have that much space between them. so the air didn't cool them all properly and some of the boards reached critical temperature.

in case you're wondering i also put the servers really close to each other and tightened as much as possible to reduce the amount of wasted air in the tunnel. kinda tried to force the air to go through the servers by not providing it much space to get out of the tunnel, other than through the servers!

I don't know maybe I did something wrong but that didn't work out for me. the length of the "wind tunnel" was 1.5 meters and had a diameter of 1.5 meters.

edit: I uploaded a picture of how dense the boards are.

 

[jrmichler]

Here's how I would tackle this problem. I'm assuming a submerged oil system, with the heat transferred to air.

1) Heat generated = 21 kW X 3412 = 72,000 BTUH (sorry about the units, but that's what I'm accustomed to).
2) Maximum temperature of oil = 65 C = 149 deg F. This is the temperature that the oil is allowed to reach on a hot (40 C) day.
3) Temperature difference between oil in and oil out: 10 deg F. This is an assumed number, you can adjust it later.
4) Oil flow rate = 72,000 BTUH / 10 deg F delta T / 7.67 lbs/gallon = 39 GPM.
5) Temperature difference between air in and air out of heat exchanger: 10 deg F. This is an assumed number, you can adjust it later.
5) Minimum air flow rate = 72,000 BTUH / 10 deg F delta T / 0.018 BTU/ft^3-deg F / 60 min/hr = 6700 CFM.
6) Decision time. You can transfer heat from oil to air by a spray system in a cooling tower, or by an oil to air heat exchanger. So evaluate both alternatives until one is clearly better.

6a) Cooling tower: My Chemical Engineers Handbook by Perry and Chilton has some good information on cooling towers. Be careful, they assume water to air which includes evaporative cooling. Also, an oil to air cooling tower will collect dirt, so you will need a good filter system. The flow rates calculated above are minimums for the assumed temperature differences. The cooling tower will need to be large enough to exchange that amount of heat without blowing an oil mist over the entire neighborhood.

6b) Heat exchanger. The above assumptions call for a heat exchanger designed for mineral oil at a flow rate of 39 GPM, 149 deg F oil in, 139 deg F oil out, 104 deg F (40 C) air in, 114 deg F air out, and 6700 CFM air flow. This could be about four automobile radiators in a square with a 36 inch panel fan on top. A 36" panel fan with a 1/2 hp motor running 870 RPM would move the required air with little noise. Aerovent is only one of many companies that make these fans: https://www.aerovent.com/wp-content...RF-BPRV-DDP-DDPRC-DDPRF-DDPRV-Catalog-168.pdf. Consider also an engineered systems such as shown by @russ_watters in Post #27.

7) This is where you start to iterate. Try a larger temperature difference on the oil to reduce oil flow rate, a larger temperature difference on air to reduce air flow rate. Either or both of those changes will affect the size of the cooling tower or heat exchanger.

(MTA) 8) The above makes assumptions about the size of the heat exchanger or size of cooling tower. If the heat exchanger is larger, you can get less temperature difference between the oil and the outside air, which would allow you to use smaller air and/or oil flow rates. Lots of iteration to get a complete solution, or take the "easy" way and use several automobile radiators, then blank them off one at a time until you find the minimum that gets the job done.

 

[Tom.G]

I'm not sure if this is repetitive, too many ideas presented above.

A minor variation to what @jrmichler suggested is put those radiators (with the oil circulating in them) in the cold pool at the bottom of the cooling towers.

At the cost of higher operating expense but smaller cooling towers, put a heat pump/air conditioner between the oil and the cooling towers. The higher water temp in the towers will increase their efficiency.

p.s. (Mentors, feel free to delete if this post is repetitive)

 

[jrmichler]

too many ideas presented above

Or possibly not enough ideas. There are many ways to solve the problem of getting rid of excess heat. Choosing the best way to solve a particular problem requires taking into account a number of factors. The amount of heat in kW, the outside air temperature, space available, allowable noise, up front cost, operating cost, reliability, durability, appearance all need to be considered.

Some of these are calculated, others are estimated or are a matter of opinion. All need to be considered.

A key point is that the OP wants to get rid of 72,000 BTUH. That's a lot of heat. It's too much heat to get rid of by adding a few computer cooling fans to blow the heat around the room. For comparison, my house needs 13,000 BTUH to maintain 72 deg F (22 C) inside when it's -20 deg F (-29 C)outside.

 

[Baluncore]

The optimum solution will be dependent on the building that houses the servers, the external climate, and the availability of water. I am surprised there is no use for the heat.

It is not surprising there have been so many suggestions. The OP has not given a clue as to the system's environment.

 

[berkeman]

It is not surprising there have been so many suggestions. The OP has not given a clue as to the system's environment.

He did say this about mid-way down page #1 of the thread:

The weather is dry and can reach up to 40°C during summer. The building is 4 story and access to the roof is possible but it's not necessary since the hot air will be dumped into open air (through the yard).

 

[Lnewqban]

I did something similar. I created a "tunnel" using tarp; at one end I put a commercial Evaporative cooler (15000 m3/h blowing power) and at the other end I placed the servers with their fans removed. the problem was that the servers' boards are really dense and don't have that much space between them. so the air didn't cool them all properly and some of the boards reached critical temperature.
...
I don't know maybe I did something wrong but that didn't work out for me. the length of the "wind tunnel" was 1.5 meters and had a diameter of 1.5 meters.

That is a huge amount of air flow (and dust), but only a fraction of it reached the surface of the heat sources.

It is unlikely that your servers are actually and simultaneously generating the amount of heat that you have assumed.
That could be your case, but in my experience, server rooms needing more than 3.5-ton of air cooling are not typical.
You will greatly oversize your cooling solution if you go by the watts that the nameplate of each server shows.

You could calculate the actual amount of heat to be removed in two ways:
1) Measuring the number of amps entering the room in certain period of time.
2) Measuring the time needed for the cooling oil to reach certain delta temperature (keeping enough forced flow of oil in contact with the heat generating surfaces is very important).

Performing several tests at the peak demand and worst environmental conditions will give you a solid number to work with.
With that number in hand, you will be able to present the problem to your superiors and to make better decisions about the more economic and practical solution.

Either you will need to resort to refrigeration (air cooled oil chiller recommended) or to evaporative cooling will make a significant financial difference.
Don't fight this problem alone, involve the persons in charge of making financial decisions and specialized technical help able to design an efficient solution.