The electricity and water use required by data centres is becoming cause for concern. Iowa and Ireland are calling for moratoriums on new development projects. Microsoft’s global water consumption grew 34% from 2021 to 2022. And estimates say that around 80% of content in data centres could be stuff we never use again, stored by default to consume energy in perpetuity for no purpose at all.
@axbom
> The electricity and water use required by data centres is becoming cause for concern
Finally. Maybe now people will realise we need to stop sending HTML by email, and using JavaScript to make website dozens of times more resource-hungry than they need to be?
> How? You can't "consume" water in the sense you consume energy. Why can't they just reuse the same water?
Data centres consume water directly for cooling, in some cases 57% sourced from potable water, and indirectly through the water requirements of non-renewable electricity generation.
Water used for cooling evaporates, and data centres do not have systems in place to capture that. Water is too cheap. Groundwater depletion is a thing, and data centres compete with farming and with local communities.
@strypey@axbom CPUs and RAM are worse but HDs also add a large part to power consumption and heat. Unlike in your notebook, server disks never turn idle.
A hierarchical file system (HFS), where unused data gets moved out to a huge tape storage (and moved back automatically when a file is accesed), actually consumes very little energy for unused data. These HFS systems are standard in scientific data centers with Exabyte-sized archives (e.g. weather forecasting and climate research).
@axbom@strypey
"Water used for cooling evaporates, and data centres do not have systems in place to capture that."
👍
I was confused how data centers "used up" water for a long time, because I assumed a closed loop like liquid-cooling a desktop.
@jbiserkov@axbom@strypey Could they cool the water by burying long loops of pipes, like a ground source heat pump? Essentially radiating it back into the ground? My suspicion is "yes, but at greater expense than wasting water they can get cheaply." So many businesses seem to exist only by shifting the costs of externalities onto the public.
@axbom
> Water used for cooling evaporates, and data centres do not have systems in place to capture that
The problem is the solution. Evaporating water to make steam is what nuclear power plants do. Hooking up the steam to the same sort of turbines could help power datacentres. The system could then condense and reuse the water.
> Iowa and Ireland are calling for moratoriums on new [datacentres]
Better to pass laws obliging them to reuse the same water or pay for it.
@strypey@axbom@jbiserkov@becha Sorry, the 2nd law of thermodynamics has some objections here. The temperature difference is far too small to power anything.
If the data centers would try to re-condensate the water, they would just get back the heat that they needed to get rid off. Plus additional heat from the energy that was used to run the pumps of the cooling system. So even worse than running w/o cooling at all.
@dgfeist
> The temperature difference is far too small to power anything
I'm going to need some details to understand your objection. Is the water evaporating or not? If it is, doesn't that mean it's making steam? Am I wrong that nuclear power plants work by making steam to turn turbines?
What am I missing?
In case it's not obvious, I'm not saying that all the energy heating the computers can be recovered this way. But that some of it could.
@dgfeist
> If the data centers would try to re-condensate the water, they would just get back the heat that they needed to get rid off
They don't need to get rid of the heat. Just to move it away from the computers.
In my mind, running the steam through turbines would harvest some of the energy keeping it above boiling point, so done right, that would condense it. Wouldn't it?
@strypey@axbom@jbiserkov@becha You are missing that the usable energy (work) you can extract from heat does not depend on the amount of heat itself but rather on the temperature difference you can achieve. The formula for the maximum theoretical efficiency e is
e = 1- T_h/T_c
T_h is the max temperature in the process (temp of your cooling air/water), T_c is outside temperature in this case. Both temps must be in Kelvin!
@becha
"Commenting on the inadequacy of dislocated, parochial attempts to challenge capitalism, Jodi Dean famously remarked: 'Goldman Sachs doesn’t care if you raise chickens'. Relocating this sentiment to the capital-driven digitalisation of the world, we can ask: 'Does Amazon care how you heat swimming pools?'."
This illustrates an attitude one might call "totalism". Either a proposal must immediately solve all problems, or it's no solution at all.
I find totalism frustrating and unhelpful, for two reasons.
One, it's a convenient excuse to do nothing now, and wait for a better solution to appear tomorrow. A classic example the perfect as the enemy of the good.
But more importantly, it ignores the way problem solving actually happens. No technological change ever gets rolled out on a mass scale without first prototyping it, and testing it on a small scale. Nor should it, due to the LoUC.
@dgfeist@strypey@axbom@jbiserkov@becha Several things are true. Most datacenters are bad at sustainable power and water use. We are right to be concerned and to push for more and faster action. The main cloud providers have the most efficient and sustainable data centers, have been working on it for many years, are publishing metrics and making progress. Amazon has several datacenters where waste heat is piped for industrial use.
@adrianco
> Most datacenters are bad at sustainable power and water use. We are right to... push for more and faster action. The main cloud providers have the most efficient and sustainable data centers, have been working on it for many years, are publishing metrics and making progress
Have they been allowing their engineers to publish documentation of their efforts, so they can collaborate across companies, bringing faster progress across the industry?
Shower, radiator or kitchen sink etc warm water in our household is heated by residual warm water of the local powerplant. Would the same with a data center be feasible ?
@ik@axbom@strypey@jbiserkov@becha The point is that power plants reach temperature differences on the order of more than 1000 °C. This is more than enough to produce high-pressure steam which can run turbines and still provide usable excess heat.
Temperature difference in a data center cooling system is probably around 30-40 °C. That is not even enough for boiling water, just for evaporation at atmospheric pressure.
@dgfeist
> Temperature difference in a data center cooling system is probably around 30-40 °C. That is not even enough for boiling water, just for evaporation at atmospheric pressure
Ah-hah! This looks like the answer to the question I asked here about what I'm missing;
@strypey@ik@axbom@jbiserkov@becha Great! In principle: at these small temperature differences, you cannot expect to extract more than ~2-3 % from the heat energy as electricity. It is better to use the heat somewhere else, e.g. for building heating.
@strypey@ik@axbom@jbiserkov@becha@dgfeist
> Makes sense. Would it be possible to make server computers that run hot enough to actually boil the cooling water, without damaging them?
No. The circuits cannot sustain such temperatures. Besides, reducing the power consumption of IT hardware also reduces the operating temperature. This is more effective.
> ...or water heating for showers and hand washing? Using it for space heating would only work for about half the year.
If you use more energy and produce more heat from the beginning, you will produce higher losses. There is no way of gaining more energy from these higher losses than from using less energy and producing less heat in the 1st place. So circuits that do not produce heat would be ideal However, this is impossible at room temperature (3rd law of thermodynamics).
@dgfeist
I think your obsession with entropy is preventing you from engaging with what I'm actually proposing in the context of the thread. Thanks for the discussion.
@axbom iowa has very little ground water. Drinking water is sourced from rivers and thus the worry about running out in droughts. sewage is treated and put back in the river (it would be safe to drink but that grosses people out) so it doesn't use water up just diverts it.
the utility that serves des moines claims tothave generated 103% of all used electric from wind. This makes iowa a natural place any data center if you want to claim green. However water in iowa is limited despite a lot of rain.
@axbom wir transformieren die reale Welt in eine digitale Welt, weil wir das können. Vielleicht ist die Menschheit eine Art schwarzes Loch für die Erde. Wir helfen dem Universum zu kollabieren, damit es in der nächsten Schleife wieder expandieren kann.
Add comment