Yet another argument to price carbon emissions. It will not only not destroy the economy, it will unleash innovation and propel us into the bright future.
(And create fun, intellectually challenging jobs for us engineers)
"A carbon tax offers the most cost-effective lever to reduce carbon emissions" is what 27 Nobel laureate economists and about 4000 other economists said in a statement in January 2019 [1]. There are some opposing opinions on some details [2]. Anyway, to an outsider this appears to be the scientific consensus.
Having a carbon tax would require to measure carbon output somewhere, isn't it? How would that work? I mean, where in the chain does the measure take place? How is it accounted?
For some reason carbon tax seems like a lot of extra and complicated steps to just tax fossil fuels way more. If we tax oil, coal and gas a lot, that's basically a simple way of taxing carbon emissions, isn't it? It's way simpler to manage, and actually sound like something realistic.
The imprecise part would be estimating and placing import taxes for products and services from countries that do not take part in the carbon tax system.
> The imprecise part would be estimating and placing import taxes for products and services from countries that do not take part in the carbon tax system.
And countries that would pretend to tax carbon but "not try that hard" to stop illegal untaxed fossil fuel use.
I'm not sure how pricing carbon dioxide emission will change datacenter operations given that Google, Facebook, Amazon, Microsoft, and Apple all have serious carbon dioxide reduction targets and all already by as much co2-free electricity as they consume.
It is difficult to know how much of that is currently cosmetic and how much is real. An actual carbon tax on fossil fuels would make everything that utilizes fossil fuels more expensive. Building materials, transporting materials by trucks, people commuting by cars, generating backup electricity for those moments when wind or solar energy do not momentarily produce enough.
> all have serious carbon dioxide reduction targets
They will drop their targets the moment the competition intensifies and threatens their profits. Voluntary commitments do not work. We have learned it the hard way over the last 30 years. Paris accord is a primary example of that.
Data centres account for about 0.3% of US emissions (less in most countries). So unless you can cut at least 200 times more co2 in other places as well, why bother?
Because other things achieve something. Me getting a new job won't save the world, but it will improve my income. This achieves nothing at all, for me or for the environment.
We have a huge problem at the moment where people think they're achieving something by personally cutting their co2 emissions by a few kg per year. Global emissions continue to rise. So they're not achieving anything.
Before we can achieve something, we have to get people to realise they are not already solving the issue.
Nobody is saying that improving data center efficiency will save the world. But saying that it achieves nothing at all is another extreme.
> people think they're achieving something by personally cutting their co2 emissions by a few kg per year.
If they reduce their emissions a bit and think that's enough - well, that's bad. If they realise it wasn't enough and more needs to be done - well, that's better than nothing. At least it's a step in the right direction.
I agree with you that global emissions just stubbornly continue to rise. Whatever we do to cut emissions just doesn't seem to matter somehow. And it's very frustrating. Yet, I think, just leaning back and doing nothing is not a solution.
I agree that doing nothing isn't a solution. But that doesn't make this a solution either. That's sort of my wider point here. We need an actual solution, this isn't that, this might be part of that, but until we have that, it's not worth anything.
I agree that we need an actual, real, global solution. But where will it come from? It's not like some benevolent aliens will write up a grand plan and hand it down to us. No. The final solution will consist of a myriad of sub solutions and measures. And improving data center efficiency will be a (albite a small one) part of it. So it's nice to have it (and other parts) at hand ready. You know, just in case the next congress gets serious about tackling the climate crisis.
The issue I have is that we have had sub-solutions for decades. We've always lacked the overall strategy and will. More subsolutions is not useful because it won't create an overall strategy.
Were in "1000 spoons when all you need is a knife" and we're rapidly searching for more spoons and celebrating spoon production.
They're actually worse than useless. People mistake more options for action. And people spend more time arguing over which sub solutions to implimented while implimenting nothing.
> We have a huge problem at the moment where people think they're achieving something by personally cutting their co2 emissions by a few kg per year.
I used to think that. Now I believe that my behavior influences other people’s behavior, and that when people make these individual behavior changes they are precursors to those individuals also doing things like getting more politically involved with environmental causes, etc.
Small changes preclude bigger ones when it comes to human behavior (I think?).
Basically the individual actions can help shift social norms and that is important, is what I now believe.
I think influence is 100 times better than individual action. But I also think the same logic applies: if it hasn't worked for 50 years, and things have gotten much much worse, why will it suddenly succeed in the next 10?
Each marginal ton of CO2 is going to make things worse. It's not like there's a threshold over which things are going to be maximally bad. It could still be worse if we pollute more.
Right now, civilisation will be over in a few generations due to climate change. If I cut my emissions to zero, and 1m other people do the same today then... civilisation will be over in a few generations.
This is what no one seems to get. The change we want is not small, we don't actually care about about delaying this a year or two. It's avert disaster or don't.
We cant solve this problem one tonne at a time. We either all get together and solve it. Or we don't solve it at all.
No one is suggesting it be solved one ton at a time. A billion different solutions all working together is how it is solved. This is one of those billion.
Exactly. I see the all or nothing, now or never argument being made very often. The 'all now, all together' is quite unlikely to happen so that leaves resigning to a 'never gonna happen' defeatism. Instead every little bit still counts, and we need to speed up contributing them to our efforts.
That's fine, but now you need another 999,999,999 solutions. Until they're here and implimented along with this one, we don't have an actual overall solution do we?
Oh wait, no, emissions went up again while we've been talking and getting this ready. So we actually need one billion and three more solutions to get to the same place the original billion would have got is to.
This is the reality of it situation.
Sorry to be so bleak. But I think people (a) think we're making progress, and by any actual measure we are not and (b) think that piecemeal, you-do-you solutions will work, when they haven't for the last 50+ years.
That's what we've done for 40 years. It's failed every single year. It's failed disastrously badly.
We need accept this has failed in order to change to something that might work. At this point, even doing nothing would be better, at least it would save time and be honest.
I am not a fan of such comments, they always want either a perfect solution or no solution at all.
There was this guy who debated with me online about using solar panels because 'they are not efficient enough' and another one who debated planting trees.
It is a step by step process. We will not magically cure up the ecology in one day in one swoosh. It'll take small efforts from various aspects.
I do not agree with this. Renewables and EV uptake is accelerating at an unprecedented pace, we just haven't hit the tipping point yet.
We're definitely not moving fast enough, but we're still moving pretty fast on the economic proposition of renewables and electrified transportation alone.
It surely doesn't help that we have leaders that continue to promote fossil fuels or unsustainable exploitation of rainforests, etc. Also unhelpful is rampant capitalism where environmental costs are not part of the equation when making profits.
But in this same system we are nearing a tipping point where renewables are the better choice for the bottom line. And that is progress.
I am quite sympathic to your predictions, I both hope and semi expect some big changes in the next decade. But I have to be strict: this is the potential for progress, not actual progress. And I think it's probably already too late, it might be too late now even to avert the worst of climate change. Adding another decade to restructure socially and get non-western non-democratic countries into the same mold is a huge delay at a very late stage.
Also, according to a BBC article, Western countries' population is going to drop when we hit 2100. It'll be interesting to see if that turns out to be true, I mean, it is a 80yr into the future prediction.
Japan's population is expected to be half. Spain and others are going in the same way, China and India are expected to reduce but not by half.
Let's hope that we keep working on this problem.
As EVs rise, I expect this to be a positive turnout. it doesn't matter that 1 EV save a lot of CO2, buses and planes are becoming electric. The more electrification we do, the more solar and wind we use, the better.
India has few hundred MW solar panel project. Other countries are also planning them, right? So let's hope that they do work in the intended way..
And a secret lag, one that no scientist, economist, politician or other expert has predicted or discovered. One effecting the whole globe.
And that huge lag cause would now need to suddenly stop causing lag (otherwise we're fucked).
And if you have all these factors, what would that mean? It would mean I was right, no point in further action, global warming was fixed around 1990 and no one noticed because of the lag!
In my opinion the energy problem is a quantitative problem and most discussions are too superficial. I want to look at grand plans, rather than specific tech since one or even two won’t be enough. We need to look at whole energy portfolios and work towards that.
> Data centers consume about 1.5 percent of the world’s electricity and are responsible for about 0.5 percent of carbon emissions. And the Internet overall is reducing greenhouse gas emissions because it distributes goods digitally that once were delivered physically, like books, music, publications and mail.
The number from the first sentence is misleading. The internet consumes about 4 percent of the world's electricity[0], data centers are just one part of it. Citation needed for the second sentence. Americans own about 25 electronic devices per household[1], and that wouldn't be the case without internet. We now stream music from YouTube all day long instead of putting one CD on repeat. A kindle costs 170kgCO2e to manufacture, whereas a second-hand book close to 0. The carbon footprint associated with the digital revolution is not clear cut because of all those behaviour changes.
IF we used new technology only to gain efficiency, and not to consume more without energy gains, this would hold true. But society is not doing that.
A second hand kindle is also close to zero emissions. Why compare a new kindle to a used book? A kindle is more a replacement for a bookshelf, and each book within has trivially low emissions
>A second hand kindle is also close to zero emissions.
Some (many?) buyers gets the latest Kindle only because they can resell the old one at a good price to second-hand buyers. Can we hold first buyers responsible for all the emissions when others partake in the product life that require new products to be created in the first place?
I wonder how much improving the efficiency of software would help --- a lot of "modern" development stacks use significantly more resources to do the same thing the older ones did, seemingly negating improvements in hardware efficiency.
The greenest instruction is the one you don't run in the first place. Remember that the next time your reason to be lazy about perf optimization is "just throw more hardware at it."
Is that true? I thought cores slept until they were needed. Or didn't clock as high if instruction load wasn't there to match.
There is of course always some nominal minimum power consumption for a server to be on. But every instruction requires some additional amount of electron flow to execute.
You can configure it to do that, though. I don't run my hardware in a datacenter (I prefer hosting from home for various reasons) but my server's CPU is set to use the powersave governor.
Let me put it a different way. There may exist two different programs that draw the same amount of power through the CPU but have different instructions-per-clock performance. The one with the higher IPC is getting all its extra instructions for free in energy terms.
Look at it another way: if a machine is already in its highest power state, with all cores running, but only executing one thread per core, running another thread on each core may cost almost nothing in terms of energy but could boost instructions retired by a lot.
The marginal cost of each instruction in compressed at the high end of utilization. The first instruction costs a lot, but after you're already running it the best thing you can do for efficiency is take utilization on that machine straight to 100%.
But if you have less intructions to run, you won't turn the machine on in the first place.
And even if the machine is already on, giving it instructions to run will make it change the CPU frequency (DVFS), access the RAM, etc. which does increase its power usage.
"access the RAM" is a big one that people overlook. Yes it might cost a few picojoules to load a line into cache, but the major cost of RAM is just that it exists and there is no low-power mode for refreshing it. RAM is incompressible and statically provisioned so if you can use memory more efficiently you can just buy less and this will save you a lot of power 24x7.
Or is the number of people in fixed supply with close to fixed CO2 output and more optimized source code just means less source code, or more people optimizing source code and less people working in the hardware supply chain, or something else along those lines.
Edit: To be clear, this is a real question. I've been mentally debating for quite awhile whether or not I should allocate workers CO2 output to companies or not.
It's an interesting question. Perhaps if you use fewer developers, a few humans will have another professional occupation which may reduce their income, and their CO2, or perhaps hiring more developers to do unnecessary tasks prevent a few humans to have jobs that produce more CO2. I feel like it's very difficult to model.
I also think that workers should be responsible for the CO2. Because otherwise, who is responsible?
Recent news from Norway claims that data centres could increase the power consumption of the country by 20%. Part of the interest is due to the clean energy and heavy reliance on hydropower in Norway, plus the cold weather conditions in the North which makes cooling somewhat easier and cheaper.^[1] At the same time parliamentary politician Lars Haltbrekken (SV)^[2] demands that such development assumes good utilization of residual heat.^[3]
As far as my own political sources claim, there is however little interest in cryptocurrency farms. To that end there is indeed rather some aversion to it, especially from the Conservative Right. On the other hand, if the Norwegian power grid is expanded for this interest, like it seems to be, I think also crypto farms would be interested as well.
You will note that none of the big league data center operators have a serious facility in California, where as you have implied energy and water are both expensive.
73 comments
[ 2.0 ms ] story [ 144 ms ] thread(And create fun, intellectually challenging jobs for us engineers)
"A carbon tax offers the most cost-effective lever to reduce carbon emissions" is what 27 Nobel laureate economists and about 4000 other economists said in a statement in January 2019 [1]. There are some opposing opinions on some details [2]. Anyway, to an outsider this appears to be the scientific consensus.
[1] https://www.econstatement.org/
[2] http://standupeconomist.com/why-im-not-signing-the-economist...
For some reason carbon tax seems like a lot of extra and complicated steps to just tax fossil fuels way more. If we tax oil, coal and gas a lot, that's basically a simple way of taxing carbon emissions, isn't it? It's way simpler to manage, and actually sound like something realistic.
This is what carbon tax means. Taxing fuels according to their carbon content. Carbon content is easy to determine precisely.
https://en.wikipedia.org/wiki/Carbon_tax
The imprecise part would be estimating and placing import taxes for products and services from countries that do not take part in the carbon tax system.
And countries that would pretend to tax carbon but "not try that hard" to stop illegal untaxed fossil fuel use.
They will drop their targets the moment the competition intensifies and threatens their profits. Voluntary commitments do not work. We have learned it the hard way over the last 30 years. Paris accord is a primary example of that.
We have a huge problem at the moment where people think they're achieving something by personally cutting their co2 emissions by a few kg per year. Global emissions continue to rise. So they're not achieving anything.
Before we can achieve something, we have to get people to realise they are not already solving the issue.
> people think they're achieving something by personally cutting their co2 emissions by a few kg per year.
If they reduce their emissions a bit and think that's enough - well, that's bad. If they realise it wasn't enough and more needs to be done - well, that's better than nothing. At least it's a step in the right direction.
I agree with you that global emissions just stubbornly continue to rise. Whatever we do to cut emissions just doesn't seem to matter somehow. And it's very frustrating. Yet, I think, just leaning back and doing nothing is not a solution.
Were in "1000 spoons when all you need is a knife" and we're rapidly searching for more spoons and celebrating spoon production.
They're actually worse than useless. People mistake more options for action. And people spend more time arguing over which sub solutions to implimented while implimenting nothing.
I used to think that. Now I believe that my behavior influences other people’s behavior, and that when people make these individual behavior changes they are precursors to those individuals also doing things like getting more politically involved with environmental causes, etc.
Small changes preclude bigger ones when it comes to human behavior (I think?).
Basically the individual actions can help shift social norms and that is important, is what I now believe.
Or: why should I do my part when everyone's selfish?
It might be difficult to measure a single such cut, but a few hundred could be noticeable. https://arstechnica.com/science/2020/07/manage-your-expectat...
This is what no one seems to get. The change we want is not small, we don't actually care about about delaying this a year or two. It's avert disaster or don't.
We cant solve this problem one tonne at a time. We either all get together and solve it. Or we don't solve it at all.
Oh wait, no, emissions went up again while we've been talking and getting this ready. So we actually need one billion and three more solutions to get to the same place the original billion would have got is to.
This is the reality of it situation.
Sorry to be so bleak. But I think people (a) think we're making progress, and by any actual measure we are not and (b) think that piecemeal, you-do-you solutions will work, when they haven't for the last 50+ years.
That's what we've done for 40 years. It's failed every single year. It's failed disastrously badly.
We need accept this has failed in order to change to something that might work. At this point, even doing nothing would be better, at least it would save time and be honest.
There was this guy who debated with me online about using solar panels because 'they are not efficient enough' and another one who debated planting trees.
It is a step by step process. We will not magically cure up the ecology in one day in one swoosh. It'll take small efforts from various aspects.
We're definitely not moving fast enough, but we're still moving pretty fast on the economic proposition of renewables and electrified transportation alone.
Right now, We're not just emitting co2, we're emitting more this year than last year.
And the amount more is rising as well.
Any actual measure of co2E emissions shows this problem is bad, its getting worse and its getting worse faster over time.
But in this same system we are nearing a tipping point where renewables are the better choice for the bottom line. And that is progress.
We will see.
Japan's population is expected to be half. Spain and others are going in the same way, China and India are expected to reduce but not by half.
Let's hope that we keep working on this problem.
As EVs rise, I expect this to be a positive turnout. it doesn't matter that 1 EV save a lot of CO2, buses and planes are becoming electric. The more electrification we do, the more solar and wind we use, the better.
India has few hundred MW solar panel project. Other countries are also planning them, right? So let's hope that they do work in the intended way..
And a secret lag, one that no scientist, economist, politician or other expert has predicted or discovered. One effecting the whole globe.
And that huge lag cause would now need to suddenly stop causing lag (otherwise we're fucked).
And if you have all these factors, what would that mean? It would mean I was right, no point in further action, global warming was fixed around 1990 and no one noticed because of the lag!
A good analysis should look like this: https://www.withouthotair.com/
The number from the first sentence is misleading. The internet consumes about 4 percent of the world's electricity[0], data centers are just one part of it. Citation needed for the second sentence. Americans own about 25 electronic devices per household[1], and that wouldn't be the case without internet. We now stream music from YouTube all day long instead of putting one CD on repeat. A kindle costs 170kgCO2e to manufacture, whereas a second-hand book close to 0. The carbon footprint associated with the digital revolution is not clear cut because of all those behaviour changes.
IF we used new technology only to gain efficiency, and not to consume more without energy gains, this would hold true. But society is not doing that.
[0] https://theshiftproject.org/en/article/unsustainable-use-onl... [1] https://www.forbes.com/sites/tjmccue/2013/01/02/24-electroni...
How is it misleading? It says "Data centers consume about 1.5 percent of the world’s electricity". It does not say anything about the internet.
Some (many?) buyers gets the latest Kindle only because they can resell the old one at a good price to second-hand buyers. Can we hold first buyers responsible for all the emissions when others partake in the product life that require new products to be created in the first place?
There is of course always some nominal minimum power consumption for a server to be on. But every instruction requires some additional amount of electron flow to execute.
Look at it another way: if a machine is already in its highest power state, with all cores running, but only executing one thread per core, running another thread on each core may cost almost nothing in terms of energy but could boost instructions retired by a lot.
The marginal cost of each instruction in compressed at the high end of utilization. The first instruction costs a lot, but after you're already running it the best thing you can do for efficiency is take utilization on that machine straight to 100%.
And even if the machine is already on, giving it instructions to run will make it change the CPU frequency (DVFS), access the RAM, etc. which does increase its power usage.
Or is the number of people in fixed supply with close to fixed CO2 output and more optimized source code just means less source code, or more people optimizing source code and less people working in the hardware supply chain, or something else along those lines.
Edit: To be clear, this is a real question. I've been mentally debating for quite awhile whether or not I should allocate workers CO2 output to companies or not.
I also think that workers should be responsible for the CO2. Because otherwise, who is responsible?
We’re still building them globally to help give electricity to 7 billion people.
All these little ideas to save a fractional percent is like the drunk looking for his keys under the light.
Decade after decade of using coal like this is a much bigger problem.
All those CO2 emissions have now cost us much more time than saving 1% in 2030
Many companies like Apple and Google are using green energy for their data centers.
Solving the electricity problem in general, solves a much bigger problem.
As far as my own political sources claim, there is however little interest in cryptocurrency farms. To that end there is indeed rather some aversion to it, especially from the Conservative Right. On the other hand, if the Norwegian power grid is expanded for this interest, like it seems to be, I think also crypto farms would be interested as well.
[1]: Norwegian language source: https://www.digi.no/artikler/star-i-ko-for-a-koble-seg-pa-da...
[2]: https://en.wikipedia.org/wiki/Socialist_Left_Party_(Norway)
[3]: Norwegian language source: https://www.nrk.no/vestland/datasentera-kan-auka-landets-kra...