Google's new Finland data center is cooled by seawater

By on May 25, 2011, 12:18 PM

Google has released a video showing off an interesting environmentally friendly innovation in one of their newly built data centers -- namely, a seawater cooling system for its servers. Located on the southern coast of Finland, the data center takes advantage of massive quarter-mile long tunnels installed by a paper mill previously operating on the site. The latter cooled some of its paper making machinery using the tunnels -- and apparently other industries have been known to do this as well -- but up until now using seawater for data center cooling was unheard of.

According to the video, the center collects cool water through an inlet pipe and travels through granite tunnels built by the paper mill in the 1950s. The water is pumped through the data center and run into exchangers that dissipate the heat from servers. Then it's mixed back with cooler water and put back into the sea at a similar temperature it entered the system, so as to have as little impact as possible on the surrounding ecosystem.

There's no air conditioning or other cooling system. The idea is part of the company’s strategy to take advantage of local resources to develop the best environmental strategy for its facilities. Next week Google will also be detailing two other energy-efficient data centers in Europe that don't use chillers, but rather outside air and evaporative cooling.

Besides minimizing environmental impact, the initiative should also help reduce operating costs, as energy consumption and cooling in particular are known to be among the biggest expenses for data centers. Apparently, the Finland data center will also make use of wind power and at least some of this would come from a new wind park next to the facility.




User Comments: 34

Got something to say? Post a comment
tacobfm said:

That's actually pretty cool.

Why not just use the colder water in the first place though?

Cueto_99 said:

Google = Innovation!

Guest said:

@ tacobfm

I'm guessing you misinterpreted "mixed back with cooler water"? It means they mixed the water used for cooling (which is now warm as a result of the process) back with cooler water (as in cooler than the now warm water, but same temperature as that water was to begin with) before pumping it back into the sea. It doesn't mean the water ejected by the system is the exact same temperature as the water going in, but the ecological impact is minimal compared to pumping out warm water straight from the cooling system.

Shnig said:

Not exactly cutting edge though. That's pretty much how Sydney Opera House's A/C works. Although they did it for vanity reasons rather then eco reasons as they didn't want coolant fans wreaking the roof curves...

Cota Cota said:

@Tacobfm

Note that due to the Conservation of Energy, that heat will eventually do the same ecological impact in some other way, just because it wont be seen next to the data center doesn't means its more "ecological"

OUTLAWXXX said:

When google isn't being scary about the data they have on you and potentially handing it over to gov. agencies... I really do love this company ^_^

captaincranky captaincranky, TechSpot Addict, said:

Seawater as a coolant? Much in the same vein, I was saying the same thing about the ten dollar hooker that just walked down the street. "Man, I'd hate to see the inside of those pipes"!...

dividebyzero dividebyzero, trainee n00b, said:

"Man, I'd hate to see the inside of those pipes"!...

Isn't that the reason that they make the maintenance inspection ducts so large?

BTW: How do you know she's $10 ? Is that the advertised MSRP ?

/not looking forward to the next Friday the 13th sale.

Lurker101 said:

dividebyzero said:

BTW: How do you know she's $10 ? Is that the advertised MSRP ?

Well you'd think the scabs would be a giveaway

Shnig said:

The reason those pipes don't rust is pretty cool. Same way some ships do it as well as the Sidney Opera House. They have a replaceable block of more reactive metal that rusts rather then the pipes. Again please everyone understand that nothing here is innovative by google in any way. Google just have more money to tell us about their "innovations" then others...

captaincranky captaincranky, TechSpot Addict, said:

The reason those pipes don't rust is pretty cool. Same way some ships do it as well as the Sidney Opera House. They have a replaceable block of more reactive metal that rusts rather then the pipes. Again please everyone understand that nothing here is innovative by google in any way. Google just have more money to tell us about their "innovations" then others...

Probably very similar in principal to a rod in your home hot water heater that collects impurities. These have been around for decades, I believe they're called, "sacrificial anodes". A description at paragraph 3 and diagram here: [link]

Guest said:

Cool man, real cool.

cmbjive said:

"Next week Google will also be detailing two other energy-efficient data centers in Europe that don't use chillers, but rather outside air and evaporative cooling."

So, they're going to use something akin to a swamp cooler to cool their servers? I wonder what they will do about inside moisture.

Also, it appears that it is still in the testing phase so we don't have an actual idea of how it will perform under pressure. Should be interesting, though.

Guest said:

IBM has used this technology in their Poughkeepsie, NY site for decades!

Guest said:

@Cota

No, it won't. That's like saying dropping a few Jell-O cubes in the middle of the pacific is going to turn the world's oceans into a gelatine desert. Ejecting warm water into the ocean is only an issue when it's concentrated and harms the natural ecosystem, if it's mixed with cold water and pumped back out it's not an issue as the amount of hot water required to be pumped into the ocean to have any effect would be massive.

Imagine taking a teaspoon of hot water, mixing it with a glass of cold water and then pouring it into an olympic swimming pool. That's essentially the process involved here, but on a much larger scale (and actually even more extreme in terms of the ratio of hot to cold water).

Jibberish18 said:

I always wondered why people near cold sources of water, whether fresh or sea, didn't use the water to cool massive server systems.

killeriii said:

@ cota

It's the lack of refrigerant that makes it more eco friendly.

Zoltan Head said:

The water is pumped through the data center and run into exchangers that dissipate the heat from servers. Then it's mixed back with cooler water and put back into the sea at a similar temperature it entered the system, so as to have as little impact as possible on the surrounding ecosystem.

The water temperature can't be very similar on return, or it isn't going to cool anything?!

dividebyzero dividebyzero, trainee n00b, said:

The water temperature can't be very similar on return, or it isn't going to cool anything?!

It's not a difficult concept, really.

cold seawater in >> water heated by components >> water partially cooled by heat exchanger/radiator (air cooling) >> cold water added to further cool flowing through the cooling loop >> cold water returned to sea

Zoltan Head said:

It's not a difficult concept, really.

cold seawater in >> water heated by components >> water partially cooled by heat exchanger/radiator (air cooling) >> cold water added to further cool flowing through the cooling loop >> cold water returned to sea

That's right, it's a simple concept, if the return temperature is similar, the cooling will be minimal, just as in the limit if the temperature is the same, no cooling will be achieved (of course the article doesn't say the return temperature is the same, just similar).

Hope this helps.

dividebyzero dividebyzero, trainee n00b, said:

Kind of depends on the heat exchanger wouldn't you think? You do realize how much heat a bank of servers puts out ?

The system I'm using to send this has a stable water-to-ambient tempreture delta of exactly 1C, with ~800ml of distilled water and ~230W heat dump (2600K @ 4.8GHz)

Sorry about the sad picture quality- it's 10pm here and the study has recessed lighting...so bad flash it is.

Zoltan Head said:

Kind of depends on the heat exchanger wouldn't you think? You do realize how much heat a bank of servers puts out ?

The system I'm using to send this has a stable water-to-ambient tempreture delta of exactly 1C, with ~800ml of distilled water and ~230W heat dump (2600K @ 4.8GHz)

Sorry about the sad picture quality- it's 10pm here and the study has recessed lighting...so bad flash it is.

It doesn't depend on the heat exchanger at all, it's just high-school Physics, there is no net heat flow between regions of equal temperature.

dividebyzero dividebyzero, trainee n00b, said:

Glad you cleared that up. I was always under the impression that thermal equilibrium tended to apply to a closed system...but since you say that it also applies to an open system it's nice to know I can do away with my watercooling and car radiators and not have to worry about a temperature rise.

I was also under the impression that for that premise to hold true, that flow in had to equal flow out- which in the case of the article isn't the case.

Thanks for the heads up. The more you know, huh!

Zoltan Head said:

Glad you cleared that up. I was always under the impression that thermal equilibrium tended to apply to a closed system...but since you say that it also applies to an open system it's nice to know I can do away with my watercooling and car radiators and not have to worry about a temperature rise.

I was also under the impression that for that premise to hold true, that flow in had to equal flow out- which in the case of the article isn't the case.

Thanks for the heads up. The more you know, huh!

What thermal equilibrium?

dividebyzero dividebyzero, trainee n00b, said:

Aren't you talking about thermodynamics ?

A quick look at a definition:

When the temperatures of interest belong to several bodies, the concept also requires that flows of heat between each pair of bodies balance to a zero net flow

and your post:

there is no net heat flow between regions of equal temperature.

I'm also quite interested to find out how :

if the return temperature is similar, the cooling will be minimal, just as in the limit if the temperature is the same, no cooling will be achieved .

Applies to my own (supposed lack) of cooling endeavours, since the article is basically a scaled up standard computer watercooling open model- non refrigerated, albeit with a secondary added coolant source and no recirculation involved.

So, for arguments sake, we use my Z77 rig as example:

Intake and outflow temperature remain constant at +1-2°C " over ambient while heat is added to the system.

Now, according to you, the heat exchanger is immaterial...

It doesn't depend on the heat exchanger at all.

...and no cooling is achieved...

no cooling will be achieved .

Despite the fact that water prior to entering the CPU block is at +1-2°C " over ambient, at +4°C " between CPU block and radiator, and +1°C " on exiting the radiator. Based on some quick and dirty calculation based on a conservative 3 litres/min flow rate, 4.18kJ/kgK specific heat for water and the temp delta across the radiator gives a ballpark average of a little over 2250 kJ/h of heat that is adding nothing to water temperature...strange, no?

LinkedKube LinkedKube, TechSpot Project Baby, said:

MEANWHILE...AT FOXCONN

LinkedKube LinkedKube, TechSpot Project Baby, said:

The water temperature can't be very similar on return, or it isn't going to cool anything?!

It's not a difficult concept, really.

cold seawater in >> water heated by components >> water partially cooled by heat exchanger/radiator (air cooling) >> cold water added to further cool flowing through the cooling loop >> cold water returned to sea

Ever pour a cold drink on an ice cube in the winter. Will you guarantee the end result will be as cold or nearly as cold as the ambient temperture?

LinkedKube LinkedKube, TechSpot Project Baby, said:

Aren't you talking about thermodynamics ?

A quick look at a definition:

When the temperatures of interest belong to several bodies, the concept also requires that flows of heat between each pair of bodies balance to a zero net flow

and your post:

there is no net heat flow between regions of equal temperature.

I'm also quite interested to find out how :

if the return temperature is similar, the cooling will be minimal, just as in the limit if the temperature is the same, no cooling will be achieved .

Applies to my own (supposed lack) of cooling endeavours, since the article is basically a scaled up standard computer watercooling open model- non refrigerated, albeit with a secondary added coolant source and no recirculation involved.

So, for arguments sake, we use my Z77 rig as example:

Intake and outflow temperature remain constant at +1-2C ? over ambient while heat is added to the system.

Now, according to you, the heat exchanger is immaterial...

It doesn't depend on the heat exchanger at all.

...and no cooling is achieved...

no cooling will be achieved .

Despite the fact that water prior to entering the CPU block is at +1-2C ? over ambient, at +4C ? between CPU block and radiator, and +1C ? on exiting the radiator. Based on some quick and dirty calculation based on a conservative 3 litres/min flow rate, 4.18kJ/kgK specific heat for water and the temp delta across the radiator gives a ballpark average of a little over 2250 kJ/h of heat that is adding nothing to water temperature...strange, no?

lmao I didn't even read this post until I noticed it was above the one I just posted.

Your post=coolant savy version of my very simple one.

Zoltan Head said:

One more go:- If there is NO difference in temperature between the thing to be cooled and the coolant, I hope it's obvious the cooler won't work, conversely if the coolant is LOTS colder than the object it will cool it very well. Then there is a continuum of situations between these extremes, at the low end of which is "if the return temperature is similar, the cooling will be minimal" mentioned earlier.

dividebyzero dividebyzero, trainee n00b, said:

One more go:- If there is NO difference in temperature between the thing to be cooled and the coolant, I hope it's obvious the cooler won't work

Well, I sincerely doubt that is an issue. The "thing" being cooled...

...is based on Google's containerized data center model. ZDNet estimate for one of these larger data centers is 650,000+ CPU cores per room. Power usage for Google's combined data center op's apparently running in excess of two-and-a-quarter billion kilowatt hours per year. With power usage pretty much equal to heat output, and the fact that the seawater coolant is taken from the Baltic Sea, I'd say it's a given that there's a pretty big temperature difference between the "thing" being cooled and the coolant.

Zoltan Head said:

I'd say it's a given that there's a pretty big temperature difference between the "thing" being cooled and the coolant.

Therefore the returning coolant must be warmer than the Sea. QED.

dividebyzero dividebyzero, trainee n00b, said:

Therefore the returning coolant must be warmer than the Sea. QED.

I would sincerely doubt that the difference is anything more than minimal, given that return coolant is already passing through a heat exchanger (in a location not that far removed from the arctic circle), a granite tunnel -a material well known for it's heatsinking ability, and is further diluted by adding further cold water. So you might say that the return water is of a similar temperature to the surrounding body of water. Which is the term used in the original article....so all that your extended posting has revealed is that you know very little about the energy requirement of a server farm, and seemingly less about liquid cooling of computers. QED.

Of course, being the pedant, feel free to search out the enviromental impact study commissioned before work started...you never know the Finnish enviroment ministry might have signed off on Google pouring warm water into the Baltic.

Zoltan Head said:

Good, agreement at last. (y)

1 person liked this | learninmypc learninmypc said:
Load all comments...

Add New Comment

TechSpot Members
Login or sign up for free,
it takes about 30 seconds.
You may also...
Get complete access to the TechSpot community. Join thousands of technology enthusiasts that contribute and share knowledge in our forum. Get a private inbox, upload your own photo gallery and more.