Slashdot Mirror
Data Centers Work To Reduce Water Usage
miller60 writes "As data centers get larger, they are getting thirstier as well. A large server farm can use up to 360,000 gallons of water a day in its cooling systems, a trend that has data center operators looking at ways to reduce their water use and impact on local water utilities. Google says two of its data centers now are "water self-sufficient." The company has built a water treatment plant at its new facility in Belgium, allowing the data center to rely on water from a nearby industrial canal. Microsoft chose San Antonio for a huge data center so it could use the local utility's recycled water ('gray water') service for the 8 million gallons it will use each month."
1. They do pay for the water
2. Why do you care what they do with water they pay for? I would not care if they used bottled water to cool their data center. The water is not destroyed, in fact it is still drinkable water, just a little warmer.
So, theoretically, through the use of evaporative cooling at large data centers, local humidity could rise, and...cloud computing could produce actual clouds?
RTFA. The water loss is because many data centers use evaporative cooling towers.
That's the point - the water does get consumed. The simplest (cheapest) way to cool the water after running it through the data center is to use evaporation towers. As the name implies, you lose a substantial portion of the water to evaporation. Evaporation towers are very efficient in terms of power and material costs, but they go through a lot of water. Costs a lot more to construct a closed-loop system - you need some sort of giant radiator to cool the water. Evap tower you just build a hollow box, put some sprayers at the top, a collector at the bottom, and off you go.
What kind of water-cooling system lets the water evaporate into the air?
Your sweat, as an example.
If I have been able to see further than others, it is because I bought a pair of binoculars.
I'm sure that they do pay something for water; but it may or may not have any relation at all to its actual cost.
For a confusing tangle of historical and political reasons, allocation of water rights is often deeply perverse. In some cases, you'll get a situation analogous to IP address (mis)allocation, where a number of entities received enormous grants of water rights many decades ago. In other cases, you'll have radically different rates across user class(frequently, agriculture ends up having access to astonishingly cheap water, compared to everybody else, and compared to the cost of producing it). In other cases, you'll have a situation where the level of water use is only maintained by sucking the aquifers dry at a rate far beyond that of replenishment, which works like a charm, up until it blows up in your face.
Because of the often dysfunctional state of pricing, uses that are flagrantly unsuitable to the location and climate often end up happening, because they don't bear anything close to the real costs of what they are doing. I can't speak for YayaY; but my concern would be not what they do with the water they pay for, they can do whatever they like, but for whether or not the price that they are paying accurately signals the cost of consuming the resource, or whether they end up imposing an externality on everybody around them.
http://en.wikipedia.org/wiki/File:Rancho_Seco_Nuclear_Generating_Station.jpg
See those big towers? Those are evaporative cooling towers. Simple, cheap, and highly efficient in terms of energy costs to operate (not so much in terms of water usage).
Ever wonder why power plants that use steam-driven generators (coal, gas, nuclear) tend to be located near large bodies of water? Same issues that high-density data center operators are discovering.
Anyone who cares about their city and it's infrastructure.
It doesn't take an environmentalist - all it takes is someone familiar with this issues who takes a moment to think.
The problem is that the water for many cities and towns comes from aquifers or dams - which rely on rain to replenish. Many of these are already highly strained, even before the load of a data center is placed on them. The water taken from these sources is then treated, which costs money, and again many cities water systems are already strained because of the high capital cost of building new ones. Again, a data center consumes so much water that this just exacerbates the problem.
I'm guessing you must not be from the US because evaporation based cooling systems are THE standard for state of the art industrial and commercial cooling in the US. If you have over 250 tons of load, you have an open cooling tower - dead standard ASHRAE design. The evaporation of water via a cooling tower is THE way you reject heat. If you want to do it dry (as is common in Europe due to much higher fear of Legionella and local code officials freaking out about it), it is FAR less efficient in almost every case, even in monsoon climates like Banglore a wet cooling tower is more efficient.
It doesn't really work for a couple reasons. First, heat doesn't get destroyed in the ground or wicked away (unless you have an underground river, which changes the whole story), it is stored. This is awesome for a building that pumps heat into the ground in the summer and then needs to pull it back out in the winter, but sucks for a datacenter that is pumping out MW of heat 8760 hours a year. Second, massive quantities of heat. A rule of thumb would be 200 feet of well per 3.5 kW of cooling. A modest datacenter is around 15 MW of waste heat, so you need 860,000 linear feet of well (with double that much piping making a U down each well). And after a year you're screwed anyhow because of issue #1.
An annoying fact of physics is that when it gets really cold, evaporative cooling becomes less effective. The air just can't hold much water, and it's the phase change from liquid to vapor that gets rid of your heat. So, it's not freezing that make low temperatures worrisome but actually loss of capacity.