As someone who really has only limited experience in data centers. I've done physical security audits for compliance purposes, I had similar thoughts when I've heard people claim DCs use water. In...
As someone who really has only limited experience in data centers. I've done physical security audits for compliance purposes, I had similar thoughts when I've heard people claim DCs use water. In all of the data centers I have audited. Which mind you is like six, and not all of them very big, all of the cooling systems have been closed-looped or conventional HVAC. Obliviously there are going to be fire suppression systems but that's not always using water. Granted I'm sure some big DCs somewhere uses some kind of liquid-cooling across racks but I'd also assume that's closed. I have a hard time thinking of some kind of solution that would continually use water inside of a DC for compute purposes.
I think people might be conflating the idea that data centers use a lot of water and chip fabs, which in my limited research uses a significant amount of water, but I have no experience or evidence to support that assumption.
Yeah, water-based systems would be rather unusual in settings like these and generally not a good idea if the thing you're trying to extinguish is mostly electronics. The datacenter at my...
Yeah, water-based systems would be rather unusual in settings like these and generally not a good idea if the thing you're trying to extinguish is mostly electronics. The datacenter at my University uses an argon-based system for example. Basically if the alarm goes off you've got a set amount of time to leave the critical areas and then the whole thing is flooded with argon gas.
Either way - if a water-based fire extinguishing system significantly contributes to your water consumption, you've usually got bigger problems than that.
I cannot tell you how many times I have found retro-fitted data centers with live water pipes running through them. Most of them at hospitals who had no other place to put them lol Flooding was...
I cannot tell you how many times I have found retro-fitted data centers with live water pipes running through them. Most of them at hospitals who had no other place to put them lol
This has been my experience as well, if anything - you absolutely do not want to use municipal or outside water in 99% of cases unless absolutely needed. For heating/cooling you can't have the...
all of the cooling systems have been closed-looped or conventional HVAC.
This has been my experience as well, if anything - you absolutely do not want to use municipal or outside water in 99% of cases unless absolutely needed. For heating/cooling you can't have the level of minerals that are needing in standard waterways, and the effect of contamination is too high and too costly not to have a closed-loop system.
That's not to say that used excess, or disposed water isn't being released into communal areas full of pollution, or that the data centers aren't using municipal water for things like restrooms, staff HVAC, etc, but that would be just standard business use levels.
Probably even less than an equivalently sized office or other business, honestly. Data centers generally donāt have a ton of employees in them at any one time, relative to their physical size.
that would be just standard business use levels.
Probably even less than an equivalently sized office or other business, honestly. Data centers generally donāt have a ton of employees in them at any one time, relative to their physical size.
If the datacenter has a closed-looped system, what would it be using more water for - let alone substantially more - than human activity in a standard office?
If the datacenter has a closed-looped system, what would it be using more water for - let alone substantially more - than human activity in a standard office?
There is no such thing as a truly closed loop system. Cooling water has to maintain a minimum degree of purity, and works best within a certain range of temperatures. Over time the water in the...
There is no such thing as a truly closed loop system. Cooling water has to maintain a minimum degree of purity, and works best within a certain range of temperatures. Over time the water in the loop will get contaminated (with microbes, salts, etc), and over time it'll also experience temperature creep (because entropy exists). So it'll have to be replaced, re-cooled, etc. There's also water loss due to evaporation, leaks, that sort of thing.
Cooling water systems at the refineries where I've worked operate in the thousands, if not millions of gallons per minute flow rates. It's not closed loop, but I'd imagine data centers probably require comparable flow rates. Let's pretend that your data center operates at a paltry 10,000 gallons per minute. If you only ever have to make up 1/100th of a percent (number out of my ass there) to maintain water balance, that's still one gallon per minute you have to maintain. Over a day, that's 1,440 gallons per day.
Let's say the typical office worker uses the toilet three times a shift, and drinks one gallon of water while at work (which would be a lot). One flush is typically 1.6 gallons of water, so per person that's roughly 6 gallons per day at the office. In order for the office to come close to the data center, it would need to have 240 people in attendance per day.
My office buildings at work don't have 240 people, so by these wild ass numbers a data center consumes more water, per day, than my office buildings do.
But again, these are rough numbers, just for food for thought.
Thanks for running the calculations and sharing your experience with similar systems. Based on your assumptions, it would indeed use a lot more water than many offices. Thankfully, It sounds like...
Thanks for running the calculations and sharing your experience with similar systems. Based on your assumptions, it would indeed use a lot more water than many offices.
Thankfully, It sounds like progress is being made on this front for data centers from what I can find. This article mentions Microsoft, Evolution Data Centres, and others are moving to "zero water" cooling.
Once the system is filled during construction, it will continually circulate water between the servers and chillers to dissipate heat without requiring a fresh water supply.
Air cooled chillers require no water during normal operation [...] These chillers use integrated compressors and condensers to cool a closed water loop. But unlike more traditional water-cooled technology, this loop is only filled once, there is no evaporative [sic] cooling and therefore no new water usage.
I think you're talking about different things without realising: the articles you mentioned are comparing closed loop to open loop/evaporative cooling systems, so the advantage of the newer...
I think you're talking about different things without realising: the articles you mentioned are comparing closed loop to open loop/evaporative cooling systems, so the advantage of the newer systems is they're "filled once" and then recirculate rather than being designed to have a continuous inflow and outflow. It's not implying they don't need to be topped up, just that they don't need the majority of the volume continually replaced.
@Drewbahr is saying that even in a closed loop system, a 0.01% loss rate adds up to a surprisingly large amount of water usage. The articles aren't about closing off that last 0.01%, they're about moving to a system that does only lose fractions of a percentage point rather than being deliberately designed to blast water in one side and out the other (which also may not actually be as bad as it sounds, depending on location and water source - you can potentially look at it as solar powered distillation if you're somewhere with abundant rain; but I wouldn't trust financial incentives to constrain its use to only the few places where that's true).
I was interested in numbers, so I found this Microsoft report comparing water use for cooling. The figures range from 2.8 - 0.02 L/kWh, and a large datacenter can draw 100MW - multiplying up and converting units gives 2,400,000 kWh/day, so that's an upper bound of about 6.7 million litres (1.8 million US gallons) per day, and a lower bound of 48,000 litres (12,700 US gallons) per day.
Given that there are plenty of datacenters significantly smaller than 100MW (an "edge" datacenter, which is more likely to be near a populated area, could easily be as little as 10MW), and we're both just doing quick back of the envelope numbers to get the order of magnitude anyway, that lower bound is pretty well in agreement with the 1,440 gallons/day estimate above.
Thank you for taking the time to try and sort things out. I did understand the claim he was making, and the uncharitable interpretation of the new proposalsā claims. Rereading the initial comment,...
Thank you for taking the time to try and sort things out. I did understand the claim he was making, and the uncharitable interpretation of the new proposalsā claims.
Rereading the initial comment, Iām not even sure theyāre talking about closed loop, so I donāt know if the 0.01% loss estimate is reasonable. Their direct experience sounds like its not with closed loop. They brought up evaporation loss, which I donāt understand how it would be an issue in a sealed closed loop system. Not sure where contamination with salts or increasing mineral content come from without evaporation and exposure. Maybe microbes could cause a biofilm buildup somewhere after a while if it was not sterilized or otherwise accounted for before sealing the system. Iām not sure why water would have to be removed/replaced for re-chilling instead of letting the systems chillers handle. Leaks I can see as a potential problem, though one would hope not significant on a new build (after initial break in) for at least a while.
At any rate, after the rest of the interaction, I have no interest in continuing conversation with them.
I'm not sure why you're taking my argument here personally, because I assure you it's not. I'm not intending to offend you; I'm just talking about physics here. My initial comment is talking about...
Exemplary
I'm not sure why you're taking my argument here personally, because I assure you it's not. I'm not intending to offend you; I'm just talking about physics here.
My initial comment is talking about closed-loop systems - ones where you don't need a continuous inflow/outflow of coolant. Modern air conditioning systems are "closed loop" in that sense, as are refrigerators and other common appliances. I'm a chemical engineer, and I've worked in industry for nearly 20 years at this point. My education and professional experience both lean quite heavily on mass and energy balance - if there's one thing a chemical engineer needs to understand, it's how energy and material moves in systems.
So when I talk about things like "evaporation loss", it's not random. Evaporation loss does happen in closed systems. Anytime you're running a coolant or refrigeration cycle, you're dealing with thermal expansion - and if you have thermal expansion, you need to have vapor space to permit the contained coolant to expand and contract. This is usually something like a reservoir, a tank, or some other "wide spot in the line". When you have a tank/reservoir, there is always some amount of coolant vapor in that space; physical laws of vapor pressure apply. When you have coolant vapor in a space like that, that is evaporation.
Now, if you want to reduce or eliminate that evaporation, you can do so - but it requires additional inputs of vapor barriers like nitrogen or air. So you can put in something like an air bladder or nitrogen blanket to inhibit the formation of this water vapor (which will never be perfect; it cannot be perfect, because physics still applies) but that creates new points where water can escape. Nitrogen blankets typically need a constant in/out flow, and if it's flowing out - so too will incremental amounts of water, which requires topping off.
Salts can develop because of incidental exposure to people - sweat, bodily fluids, that sort of thing. People need to work on these systems, and people aren't perfectly clean. If the cooling system is in a cleanroom environment you can eliminate most of these issues, but never all of them. But the concern about salts isn't limited to the interior of the cooling system; they can accumulate on the exterior as well, which can lead to corrosion and damage to piping. Again, this can be caused by people and bodily fluids, but also from animals (which probably aren't going to be a major problem, but who knows), the ambient environment, that sort of thing.
Anywhere that there is water, there are microbes. If you want to disinfect the water to limit microbial growth, you can do so - but it requires chemical addition to your water, which means adding things like bleaching agents and other halogen-containing compounds. When you add bleaching agents to systems containing metals, you'll create salts - which brings us back to the problem mentioned above. Now, you can create your cooling system out of plastic piping (PVC, HDPE, etc), but plastics and common bleaching agents don't always play well together.
Beyond all of this, there's also the problem of moving parts. For a typical cooling system, you'll need your reservoir, some pumps, and some heat exchangers (probably air-cooled). All of these - the pumps in particular - have moving parts, probably made out of metal. These parts wear out. In the case of pumps, "wearing out" will include the gradual grinding down of the gears within the pump, which will start sending tiny metal fragments throughout the system. These will promote rusting, which will not only cause damage to the piping but also create accumulation of corrosive byproducts which will lead to further corrosion, as well as affecting water purity.
As for re-chilling, when I said that physical laws apply, I'm also referring to Newton's Laws of Thermodynamics. You cannot create a perfect closed-loop that loses zero energy, it is physically impossible. You always lose some of your energy to entropy and waste heat generation from things like friction (2nd Law). Over time, your cooling system will become less cool, and will require either an influx of new, cooler coolant ... or will require re-application of external cooling to bring it back to within your pre-determined requirements.
Leaks will always, always happen - anyplace you have a weld, a joint, a flange, a change in direction, or a connection, there is a point that a leak can occur. Anyplace you have a valve for filling up, draining, installation of an instrument, that sort of thing - all of those are potential leak points.
All of these factors, and more, are why I say you cannot create a truly closed-loop system. It is simply not possible. And even beyond the impossibility of it, there's also the financial reality that while you can create really, really tight "closed-loop systems" that have near-zero leakage and loss, it is never truly zero ... and the financial costs associated with trying to make that perfect closed-loop system may be exorbitantly high. It is often far less expensive to create an open-loop, particularly for a major installation like a data center. For something small like a home air conditioning system, it's a different story; the scale is vastly different.
Something about the dismissive one-liners at the time I read them really rustled my jimmies, giving me flashbacks to the snark and arrogance of Reddit and HackerNews even if it wasn't intended...
Something about the dismissive one-liners at the time I read them really rustled my jimmies, giving me flashbacks to the snark and arrogance of Reddit and HackerNews even if it wasn't intended that way. I should have stepped away for a moment, remembered to apply charitable readings on Tildes, and asked more questions instead. I'm sorry and I'll try to do better, and I genuinely appreciate your elaboration here.
This is the bit that confused me on what kind of systems you've had experience with (emphasis mine):
Cooling water systems at the refineries where I've worked operate in the thousands, if not millions of gallons per minute flow rates. It's not closed loop, but I'd imagine data centers probably require comparable flow rates.
I read it as "I've worked with large systems. They [those systems] are not closed loop, but I'd imagine data centers operate similarly", which was the wrong interpretation.
Anyway
Evaporation within the system certainly makes sense. Wouldn't that eventually reach equilibrium and be factored into the construction, or does it still count as closed loop even if it's allowing evaporation to the atmosphere as part of the designed cooling process?
For salts and such I had only considered the obvious sources such like inflow/replacement, increasing concentrations from lossy evaporation, or harsh environments (e.g. near/in the sea). I would have never thought about sweaty maintenance workers as a significant source over time, slowly deteriorating pumps, and additives (like disinfectants) reacting with the metal parts to create salts.
Do you see the alternatives and proposals for the datacenter cooling that Microsoft and the others mentioned as being substantial improvements, getting as close as is possible to "zero-water" after construction?
Evaporation would almost certainly never reach "true" equilibrium for all of the reasons I mentioned in my previous, long post. Vapor will find ways to get out of the system, which will cause more...
Evaporation would almost certainly never reach "true" equilibrium for all of the reasons I mentioned in my previous, long post. Vapor will find ways to get out of the system, which will cause more vapor to evolve out of the liquid (see: vapor pressure). A system that allows for evaporation to the atmosphere is inherently not closed-loop, because it is losing material to the air at all times. That is, in fact, how arguably the most popular form of open-loop cooling works (evaporative cooling).
Evaporation within a closed loop needs to be limited either by air bladder/nitrogen blankets with some form of water reclamation system, or by additional cooling being applied to inhibit the formation of vapors in the first place. Either way, you're adding or removing energy in some form that needs to be accounted for. It's not impossible, it's just complex.
Regarding the datacenter cooling you mentioned previously ... I'd have more of an opinion if I actually knew what they were doing. The first article - "Sustainable by design: Next-generation datacenters consume zero water for cooling" - does not describe what they are actually doing, only that they are using less water (via a chart with no sources listed). That article reads more like an advertisement than an article describing actual measures being taken to reduce water usage.
The other article/page - "Reducing Data Centre Environmental Impact with Waterless Cooling" - only sparingly mentions ways that they would reduce water consumption, via air coolers with "integrated compressors and condensers". I have a good idea about what they mean - it's similar to the lube and seal oil systems used by machines all over the world, which are "closed loops" that do not require the constant addition of oil. What I can say is, those systems suffer from all of the issues I mentioned previously - machines break down over time, corrosive elements build up, etc etc.
In particular, the second article mentions that "Little maintenance is required for the closed loop, simply the occasional addition of water treatment to control oxidation and bacteria growth." If they're controlling oxidation, that means they're controlling rust - which likely means the addition of acids or bases to scavenge oxygen and other oxidizers that accumulate over time. Adding acids or bases means adding corrosive chemicals to your water supply; they need to be controlled with a gentle touch, and the physical orientation of the piping is going to matter a lot.
As an example of this, a facility that I worked for, years ago, had an acid vapor recovery system that needed to be replaced. They selected a steel alloy designed specifically to withstand the kind of corrosion they were experiencing in the existing piping. The problem was, their pipe racks - the supporting system keeping all of the pipes elevated off the ground, so people could work in the area - were quite crowded. Of course, the obvious answer would be to expand the pipe racks; however, that costs a lot of money and requires a lot of infrastructure to be built. So instead, they opted to remove the corroded piping and install the new alloy piping in its place.
Problem was, the piping needed to be installed as free-draining - meaning, it had to have a slope back to the acid tanks, so that any liquid that precipitated or condensed would automatically drain back into the acid tanks via gravity. That way, nothing acidic would accumulate in the piping itself, and only the vapor would flow out into the recovery system. There was one little stretch of piping that couldn't quite meet the typical 1" per 100ft of slope required for free-draining; it had to be installed dead-level. Normally, a little stretch of dead-level piping wouldn't be an issue - there's enough vapor traffic and liquid drainage that it would have no real negative effect.
So, the engineers carefully measured everything and got the piping installed. Unfortunately, it didn't quite get installed dead-level; it had a very, very slight slope away from the acid tanks, rather than toward them. The result? There was a slight accumulation of acidic precipitate in the acid gas recovery piping, and the piping - built from an alloy specifically intended to withstand the kinds of corrosion it was going to experience - suffered through-wall corrosion in less than 6 months. The whole effort had to be done again with a full replacement of the already-replaced piping, which effectively doubled the cost of the project.
Now, the kinds of acidic liquid that system was dealing with were way, way more severe than what a cooling system like we're discussing here would experience. I mention that example, though, as a way to point out that you can design a system, on paper, to be perfect for its intended use - but physical reality and constraints can conspire to ruin it quite quickly.
No worries! And for what itās worth, since you do seem interested in the mechanics at play here: closed loop doesnāt necessarily imply sealed. Most indoor swimming pools are considered closed loop...
No worries! And for what itās worth, since you do seem interested in the mechanics at play here: closed loop doesnāt necessarily imply sealed. Most indoor swimming pools are considered closed loop systems even though theyāre open to the air, for example, because the same water is continuously recirculated through.
Even when a loop is isolated from the outside environment, no seal is mechanically perfect. From the single-machine level upwards you need to consider the permeability of the tubing material to evaporation, the tightness of seals between components and at disconnection points, the possibility of loop contamination when adding/removing components for upgrades or maintenance, etc etc. and all of those concerns go up in proportion to the scale of the installation.
I know it's not the same scale, but isn't this like Watercooled desktop PC? I mean sure, you have to periodically flush and refill the loop so consumption isn't literally zero, but it's pretty...
I know it's not the same scale, but isn't this like Watercooled desktop PC? I mean sure, you have to periodically flush and refill the loop so consumption isn't literally zero, but it's pretty damn close...
A desktop PC uses several orders of magnitude less water than a data center. It's not comparable, even at scale - in no small part because the number of points of failure are substantially less in...
A desktop PC uses several orders of magnitude less water than a data center. It's not comparable, even at scale - in no small part because the number of points of failure are substantially less in a PC than for a large-scale installation.
Evaporative cooling is very common in data centers, especially very large ones. Depending on local conditions, it can use a lot less energy than refrigeration cycle cooling, with the drawback that...
Evaporative cooling is very common in data centers, especially very large ones. Depending on local conditions, it can use a lot less energy than refrigeration cycle cooling, with the drawback that you'll be constantly using water.
This was my impression as well, that the high water usage was from evaporative cooling towers which can use a tremendous amount of water. That said, this can still be an over-all water reduction...
This was my impression as well, that the high water usage was from evaporative cooling towers which can use a tremendous amount of water. That said, this can still be an over-all water reduction compared to just a standard air-cooled air-conditioning system because of the reduced electricity. At least if the electricity comes from a petroleum fuel source which also tends to need a constant supply of fresh water for cooling, so by reducing electricity requirements directly reduces the overall water requirements. Granted this is an over-all water reduction, but not necessarily water reduction from a specific water source if the electric plant and data-center rely on two different water sources. The data-center could still stress the fresh water source it's on.
Yes, exactly. People are making the argument that evaporative cooling is wasteful, but it isn't. It's more efficient from an energy use perspective than refrigeration cycles, because forcing heat...
Yes, exactly. People are making the argument that evaporative cooling is wasteful, but it isn't. It's more efficient from an energy use perspective than refrigeration cycles, because forcing heat directly work against the second law of thermodynamics requires a ton of power. Merely moving that heat somewhere else by heating up then evaporating a carrier fluid is much cheaper, and our planet is very good at cooling water vapor as part of the water cycle.
The thing that is inefficient is the data centers themselves, not the method of cooling them. You're going to use a ton of energy to keep them cool, either via electricity, or via the downstream effects of using a ton of water. The water method is usually cheaper, which usually translates to more efficient.
But depending on where you are, liquid water may be scarce and the data center use may compete directly with other uses such as drinking bathing and farming. Arizona comes to mind as an example....
But depending on where you are, liquid water may be scarce and the data center use may compete directly with other uses such as drinking bathing and farming. Arizona comes to mind as an example. So does Mexico City.
People who have never experienced government enforcement of water rationing may not appreciate the frustration and resentment that can be associated with a new industry coming in to use yet more water in a system that is already stressed.
I think that's what gets me about this. I understand the argument but by not comparing the data center use to the household water use (because it's industry) the fact is that household water use...
I think that's what gets me about this. I understand the argument but by not comparing the data center use to the household water use (because it's industry) the fact is that household water use can still be impacted by increased usage. Especially if they're being asked to restrict water use due to current drought conditions. Which is always the problem with telling people to limit showers and toilet flushes while major commercial water use seems to keep going. (And most other household environmental responses IMO) No one likes feeling like they're being restricted while a corporation is coming in and taking even more of the restricted resource.
Yes, and depending on where you live power may be scarce as well. Data centers use a ton of resources. It's either a ton of electricity, or less electricity and a lot of water. Either way they're...
Yes, and depending on where you live power may be scarce as well. Data centers use a ton of resources. It's either a ton of electricity, or less electricity and a lot of water.
Either way they're going to use a ton of resources that could be used for other things instead. Specifically what makes the most economic sense for the owner of the data center and what will have the least impact on the local area depends on the area.
Without a doubt an important consideration (that seems to be getting overlooked by some cities which is frustrating). Which the added massive electric load likely also places additional strain on...
Without a doubt an important consideration (that seems to be getting overlooked by some cities which is frustrating). Which the added massive electric load likely also places additional strain on the local water resources in a water scarce region before even factoring in the cooling on top of it.
But a lot of the politics (and big lack of apparent science driving the policies around it) in water scarce regions in the US are a complete mystery to me. Water is all the more important for the actual people living there, why incentivise companies and people who want to use it for water intensive applications to do that? š They should go somewhere without that scarcity.
I hope the water utilities are raising hell to who ever is greenlighting the projects in places where people's taps barely trickle because the pressure is so low. That becomes a health and safety issue once the pressure in the system drops since it'll start sucking in contaminated ground water back in through system leaks.
Just feels like death spiral without a clean resolution that doesn't end with a lot of frustration, expense, and likely endangered health/life.
As someone who really has only limited experience in data centers. I've done physical security audits for compliance purposes, I had similar thoughts when I've heard people claim DCs use water. In all of the data centers I have audited. Which mind you is like six, and not all of them very big, all of the cooling systems have been closed-looped or conventional HVAC. Obliviously there are going to be fire suppression systems but that's not always using water. Granted I'm sure some big DCs somewhere uses some kind of liquid-cooling across racks but I'd also assume that's closed. I have a hard time thinking of some kind of solution that would continually use water inside of a DC for compute purposes.
I think people might be conflating the idea that data centers use a lot of water and chip fabs, which in my limited research uses a significant amount of water, but I have no experience or evidence to support that assumption.
Please don't use water-based fire suppression next to my server racks š
Yeah, water-based systems would be rather unusual in settings like these and generally not a good idea if the thing you're trying to extinguish is mostly electronics. The datacenter at my University uses an argon-based system for example. Basically if the alarm goes off you've got a set amount of time to leave the critical areas and then the whole thing is flooded with argon gas.
Either way - if a water-based fire extinguishing system significantly contributes to your water consumption, you've usually got bigger problems than that.
I cannot tell you how many times I have found retro-fitted data centers with live water pipes running through them. Most of them at hospitals who had no other place to put them lol
Flooding was more common than you would think.
This has been my experience as well, if anything - you absolutely do not want to use municipal or outside water in 99% of cases unless absolutely needed. For heating/cooling you can't have the level of minerals that are needing in standard waterways, and the effect of contamination is too high and too costly not to have a closed-loop system.
That's not to say that used excess, or disposed water isn't being released into communal areas full of pollution, or that the data centers aren't using municipal water for things like restrooms, staff HVAC, etc, but that would be just standard business use levels.
Probably even less than an equivalently sized office or other business, honestly. Data centers generally donāt have a ton of employees in them at any one time, relative to their physical size.
I would imagine flushing toilets and using sinks costumes substantially less water than a data center.
If the datacenter has a closed-looped system, what would it be using more water for - let alone substantially more - than human activity in a standard office?
There is no such thing as a truly closed loop system. Cooling water has to maintain a minimum degree of purity, and works best within a certain range of temperatures. Over time the water in the loop will get contaminated (with microbes, salts, etc), and over time it'll also experience temperature creep (because entropy exists). So it'll have to be replaced, re-cooled, etc. There's also water loss due to evaporation, leaks, that sort of thing.
Cooling water systems at the refineries where I've worked operate in the thousands, if not millions of gallons per minute flow rates. It's not closed loop, but I'd imagine data centers probably require comparable flow rates. Let's pretend that your data center operates at a paltry 10,000 gallons per minute. If you only ever have to make up 1/100th of a percent (number out of my ass there) to maintain water balance, that's still one gallon per minute you have to maintain. Over a day, that's 1,440 gallons per day.
Let's say the typical office worker uses the toilet three times a shift, and drinks one gallon of water while at work (which would be a lot). One flush is typically 1.6 gallons of water, so per person that's roughly 6 gallons per day at the office. In order for the office to come close to the data center, it would need to have 240 people in attendance per day.
My office buildings at work don't have 240 people, so by these wild ass numbers a data center consumes more water, per day, than my office buildings do.
But again, these are rough numbers, just for food for thought.
Thanks for running the calculations and sharing your experience with similar systems. Based on your assumptions, it would indeed use a lot more water than many offices.
Thankfully, It sounds like progress is being made on this front for data centers from what I can find. This article mentions Microsoft, Evolution Data Centres, and others are moving to "zero water" cooling.
Microsoft info
Evolution Data Centres info in Singapore where water scarcity and high humidity can be issues
Any system that insists it will only be filled once, is lying to you. Basic laws of physics apply to everything.
Feel free to be as cynical as you want.
I think you're talking about different things without realising: the articles you mentioned are comparing closed loop to open loop/evaporative cooling systems, so the advantage of the newer systems is they're "filled once" and then recirculate rather than being designed to have a continuous inflow and outflow. It's not implying they don't need to be topped up, just that they don't need the majority of the volume continually replaced.
@Drewbahr is saying that even in a closed loop system, a 0.01% loss rate adds up to a surprisingly large amount of water usage. The articles aren't about closing off that last 0.01%, they're about moving to a system that does only lose fractions of a percentage point rather than being deliberately designed to blast water in one side and out the other (which also may not actually be as bad as it sounds, depending on location and water source - you can potentially look at it as solar powered distillation if you're somewhere with abundant rain; but I wouldn't trust financial incentives to constrain its use to only the few places where that's true).
I was interested in numbers, so I found this Microsoft report comparing water use for cooling. The figures range from 2.8 - 0.02 L/kWh, and a large datacenter can draw 100MW - multiplying up and converting units gives 2,400,000 kWh/day, so that's an upper bound of about 6.7 million litres (1.8 million US gallons) per day, and a lower bound of 48,000 litres (12,700 US gallons) per day.
Given that there are plenty of datacenters significantly smaller than 100MW (an "edge" datacenter, which is more likely to be near a populated area, could easily be as little as 10MW), and we're both just doing quick back of the envelope numbers to get the order of magnitude anyway, that lower bound is pretty well in agreement with the 1,440 gallons/day estimate above.
Thank you for taking the time to try and sort things out. I did understand the claim he was making, and the uncharitable interpretation of the new proposalsā claims.
Rereading the initial comment, Iām not even sure theyāre talking about closed loop, so I donāt know if the 0.01% loss estimate is reasonable. Their direct experience sounds like its not with closed loop. They brought up evaporation loss, which I donāt understand how it would be an issue in a sealed closed loop system. Not sure where contamination with salts or increasing mineral content come from without evaporation and exposure. Maybe microbes could cause a biofilm buildup somewhere after a while if it was not sterilized or otherwise accounted for before sealing the system. Iām not sure why water would have to be removed/replaced for re-chilling instead of letting the systems chillers handle. Leaks I can see as a potential problem, though one would hope not significant on a new build (after initial break in) for at least a while.
At any rate, after the rest of the interaction, I have no interest in continuing conversation with them.
I'm not sure why you're taking my argument here personally, because I assure you it's not. I'm not intending to offend you; I'm just talking about physics here.
My initial comment is talking about closed-loop systems - ones where you don't need a continuous inflow/outflow of coolant. Modern air conditioning systems are "closed loop" in that sense, as are refrigerators and other common appliances. I'm a chemical engineer, and I've worked in industry for nearly 20 years at this point. My education and professional experience both lean quite heavily on mass and energy balance - if there's one thing a chemical engineer needs to understand, it's how energy and material moves in systems.
So when I talk about things like "evaporation loss", it's not random. Evaporation loss does happen in closed systems. Anytime you're running a coolant or refrigeration cycle, you're dealing with thermal expansion - and if you have thermal expansion, you need to have vapor space to permit the contained coolant to expand and contract. This is usually something like a reservoir, a tank, or some other "wide spot in the line". When you have a tank/reservoir, there is always some amount of coolant vapor in that space; physical laws of vapor pressure apply. When you have coolant vapor in a space like that, that is evaporation.
Now, if you want to reduce or eliminate that evaporation, you can do so - but it requires additional inputs of vapor barriers like nitrogen or air. So you can put in something like an air bladder or nitrogen blanket to inhibit the formation of this water vapor (which will never be perfect; it cannot be perfect, because physics still applies) but that creates new points where water can escape. Nitrogen blankets typically need a constant in/out flow, and if it's flowing out - so too will incremental amounts of water, which requires topping off.
Salts can develop because of incidental exposure to people - sweat, bodily fluids, that sort of thing. People need to work on these systems, and people aren't perfectly clean. If the cooling system is in a cleanroom environment you can eliminate most of these issues, but never all of them. But the concern about salts isn't limited to the interior of the cooling system; they can accumulate on the exterior as well, which can lead to corrosion and damage to piping. Again, this can be caused by people and bodily fluids, but also from animals (which probably aren't going to be a major problem, but who knows), the ambient environment, that sort of thing.
Anywhere that there is water, there are microbes. If you want to disinfect the water to limit microbial growth, you can do so - but it requires chemical addition to your water, which means adding things like bleaching agents and other halogen-containing compounds. When you add bleaching agents to systems containing metals, you'll create salts - which brings us back to the problem mentioned above. Now, you can create your cooling system out of plastic piping (PVC, HDPE, etc), but plastics and common bleaching agents don't always play well together.
Beyond all of this, there's also the problem of moving parts. For a typical cooling system, you'll need your reservoir, some pumps, and some heat exchangers (probably air-cooled). All of these - the pumps in particular - have moving parts, probably made out of metal. These parts wear out. In the case of pumps, "wearing out" will include the gradual grinding down of the gears within the pump, which will start sending tiny metal fragments throughout the system. These will promote rusting, which will not only cause damage to the piping but also create accumulation of corrosive byproducts which will lead to further corrosion, as well as affecting water purity.
As for re-chilling, when I said that physical laws apply, I'm also referring to Newton's Laws of Thermodynamics. You cannot create a perfect closed-loop that loses zero energy, it is physically impossible. You always lose some of your energy to entropy and waste heat generation from things like friction (2nd Law). Over time, your cooling system will become less cool, and will require either an influx of new, cooler coolant ... or will require re-application of external cooling to bring it back to within your pre-determined requirements.
Leaks will always, always happen - anyplace you have a weld, a joint, a flange, a change in direction, or a connection, there is a point that a leak can occur. Anyplace you have a valve for filling up, draining, installation of an instrument, that sort of thing - all of those are potential leak points.
All of these factors, and more, are why I say you cannot create a truly closed-loop system. It is simply not possible. And even beyond the impossibility of it, there's also the financial reality that while you can create really, really tight "closed-loop systems" that have near-zero leakage and loss, it is never truly zero ... and the financial costs associated with trying to make that perfect closed-loop system may be exorbitantly high. It is often far less expensive to create an open-loop, particularly for a major installation like a data center. For something small like a home air conditioning system, it's a different story; the scale is vastly different.
Something about the dismissive one-liners at the time I read them really rustled my jimmies, giving me flashbacks to the snark and arrogance of Reddit and HackerNews even if it wasn't intended that way. I should have stepped away for a moment, remembered to apply charitable readings on Tildes, and asked more questions instead. I'm sorry and I'll try to do better, and I genuinely appreciate your elaboration here.
This is the bit that confused me on what kind of systems you've had experience with (emphasis mine):
I read it as "I've worked with large systems. They [those systems] are not closed loop, but I'd imagine data centers operate similarly", which was the wrong interpretation.
Anyway
Evaporation within the system certainly makes sense. Wouldn't that eventually reach equilibrium and be factored into the construction, or does it still count as closed loop even if it's allowing evaporation to the atmosphere as part of the designed cooling process?
For salts and such I had only considered the obvious sources such like inflow/replacement, increasing concentrations from lossy evaporation, or harsh environments (e.g. near/in the sea). I would have never thought about sweaty maintenance workers as a significant source over time, slowly deteriorating pumps, and additives (like disinfectants) reacting with the metal parts to create salts.
Do you see the alternatives and proposals for the datacenter cooling that Microsoft and the others mentioned as being substantial improvements, getting as close as is possible to "zero-water" after construction?
Evaporation would almost certainly never reach "true" equilibrium for all of the reasons I mentioned in my previous, long post. Vapor will find ways to get out of the system, which will cause more vapor to evolve out of the liquid (see: vapor pressure). A system that allows for evaporation to the atmosphere is inherently not closed-loop, because it is losing material to the air at all times. That is, in fact, how arguably the most popular form of open-loop cooling works (evaporative cooling).
Evaporation within a closed loop needs to be limited either by air bladder/nitrogen blankets with some form of water reclamation system, or by additional cooling being applied to inhibit the formation of vapors in the first place. Either way, you're adding or removing energy in some form that needs to be accounted for. It's not impossible, it's just complex.
Regarding the datacenter cooling you mentioned previously ... I'd have more of an opinion if I actually knew what they were doing. The first article - "Sustainable by design: Next-generation datacenters consume zero water for cooling" - does not describe what they are actually doing, only that they are using less water (via a chart with no sources listed). That article reads more like an advertisement than an article describing actual measures being taken to reduce water usage.
The other article/page - "Reducing Data Centre Environmental Impact with Waterless Cooling" - only sparingly mentions ways that they would reduce water consumption, via air coolers with "integrated compressors and condensers". I have a good idea about what they mean - it's similar to the lube and seal oil systems used by machines all over the world, which are "closed loops" that do not require the constant addition of oil. What I can say is, those systems suffer from all of the issues I mentioned previously - machines break down over time, corrosive elements build up, etc etc.
In particular, the second article mentions that "Little maintenance is required for the closed loop, simply the occasional addition of water treatment to control oxidation and bacteria growth." If they're controlling oxidation, that means they're controlling rust - which likely means the addition of acids or bases to scavenge oxygen and other oxidizers that accumulate over time. Adding acids or bases means adding corrosive chemicals to your water supply; they need to be controlled with a gentle touch, and the physical orientation of the piping is going to matter a lot.
As an example of this, a facility that I worked for, years ago, had an acid vapor recovery system that needed to be replaced. They selected a steel alloy designed specifically to withstand the kind of corrosion they were experiencing in the existing piping. The problem was, their pipe racks - the supporting system keeping all of the pipes elevated off the ground, so people could work in the area - were quite crowded. Of course, the obvious answer would be to expand the pipe racks; however, that costs a lot of money and requires a lot of infrastructure to be built. So instead, they opted to remove the corroded piping and install the new alloy piping in its place.
Problem was, the piping needed to be installed as free-draining - meaning, it had to have a slope back to the acid tanks, so that any liquid that precipitated or condensed would automatically drain back into the acid tanks via gravity. That way, nothing acidic would accumulate in the piping itself, and only the vapor would flow out into the recovery system. There was one little stretch of piping that couldn't quite meet the typical 1" per 100ft of slope required for free-draining; it had to be installed dead-level. Normally, a little stretch of dead-level piping wouldn't be an issue - there's enough vapor traffic and liquid drainage that it would have no real negative effect.
So, the engineers carefully measured everything and got the piping installed. Unfortunately, it didn't quite get installed dead-level; it had a very, very slight slope away from the acid tanks, rather than toward them. The result? There was a slight accumulation of acidic precipitate in the acid gas recovery piping, and the piping - built from an alloy specifically intended to withstand the kinds of corrosion it was going to experience - suffered through-wall corrosion in less than 6 months. The whole effort had to be done again with a full replacement of the already-replaced piping, which effectively doubled the cost of the project.
Now, the kinds of acidic liquid that system was dealing with were way, way more severe than what a cooling system like we're discussing here would experience. I mention that example, though, as a way to point out that you can design a system, on paper, to be perfect for its intended use - but physical reality and constraints can conspire to ruin it quite quickly.
No worries! And for what itās worth, since you do seem interested in the mechanics at play here: closed loop doesnāt necessarily imply sealed. Most indoor swimming pools are considered closed loop systems even though theyāre open to the air, for example, because the same water is continuously recirculated through.
Even when a loop is isolated from the outside environment, no seal is mechanically perfect. From the single-machine level upwards you need to consider the permeability of the tubing material to evaporation, the tightness of seals between components and at disconnection points, the possibility of loop contamination when adding/removing components for upgrades or maintenance, etc etc. and all of those concerns go up in proportion to the scale of the installation.
It's not cynicism. It's physical reality.
I know it's not the same scale, but isn't this like Watercooled desktop PC? I mean sure, you have to periodically flush and refill the loop so consumption isn't literally zero, but it's pretty damn close...
A desktop PC uses several orders of magnitude less water than a data center. It's not comparable, even at scale - in no small part because the number of points of failure are substantially less in a PC than for a large-scale installation.
Evaporative cooling is very common in data centers, especially very large ones. Depending on local conditions, it can use a lot less energy than refrigeration cycle cooling, with the drawback that you'll be constantly using water.
This was my impression as well, that the high water usage was from evaporative cooling towers which can use a tremendous amount of water. That said, this can still be an over-all water reduction compared to just a standard air-cooled air-conditioning system because of the reduced electricity. At least if the electricity comes from a petroleum fuel source which also tends to need a constant supply of fresh water for cooling, so by reducing electricity requirements directly reduces the overall water requirements. Granted this is an over-all water reduction, but not necessarily water reduction from a specific water source if the electric plant and data-center rely on two different water sources. The data-center could still stress the fresh water source it's on.
Yes, exactly. People are making the argument that evaporative cooling is wasteful, but it isn't. It's more efficient from an energy use perspective than refrigeration cycles, because forcing heat directly work against the second law of thermodynamics requires a ton of power. Merely moving that heat somewhere else by heating up then evaporating a carrier fluid is much cheaper, and our planet is very good at cooling water vapor as part of the water cycle.
The thing that is inefficient is the data centers themselves, not the method of cooling them. You're going to use a ton of energy to keep them cool, either via electricity, or via the downstream effects of using a ton of water. The water method is usually cheaper, which usually translates to more efficient.
But depending on where you are, liquid water may be scarce and the data center use may compete directly with other uses such as drinking bathing and farming. Arizona comes to mind as an example. So does Mexico City.
People who have never experienced government enforcement of water rationing may not appreciate the frustration and resentment that can be associated with a new industry coming in to use yet more water in a system that is already stressed.
I think that's what gets me about this. I understand the argument but by not comparing the data center use to the household water use (because it's industry) the fact is that household water use can still be impacted by increased usage. Especially if they're being asked to restrict water use due to current drought conditions. Which is always the problem with telling people to limit showers and toilet flushes while major commercial water use seems to keep going. (And most other household environmental responses IMO) No one likes feeling like they're being restricted while a corporation is coming in and taking even more of the restricted resource.
Yes, and depending on where you live power may be scarce as well. Data centers use a ton of resources. It's either a ton of electricity, or less electricity and a lot of water.
Either way they're going to use a ton of resources that could be used for other things instead. Specifically what makes the most economic sense for the owner of the data center and what will have the least impact on the local area depends on the area.
Without a doubt an important consideration (that seems to be getting overlooked by some cities which is frustrating). Which the added massive electric load likely also places additional strain on the local water resources in a water scarce region before even factoring in the cooling on top of it.
But a lot of the politics (and big lack of apparent science driving the policies around it) in water scarce regions in the US are a complete mystery to me. Water is all the more important for the actual people living there, why incentivise companies and people who want to use it for water intensive applications to do that? š They should go somewhere without that scarcity.
I hope the water utilities are raising hell to who ever is greenlighting the projects in places where people's taps barely trickle because the pressure is so low. That becomes a health and safety issue once the pressure in the system drops since it'll start sucking in contaminated ground water back in through system leaks.
Just feels like death spiral without a clean resolution that doesn't end with a lot of frustration, expense, and likely endangered health/life.
Why would they? The people living near data centers donāt really have running water anymore.
Can't use what you don't have. (ā āļ¾ā āļ¾ā )ā ā