How Sustainable Is a Solar Powered Website?(lowtechmagazine.com) |
How Sustainable Is a Solar Powered Website?(lowtechmagazine.com) |
The article claims that it takes "3,514 MJ of energy to produce one m2 of solar panel."
The source for that assertion is this article from 2017:
"Energy Payback Time of a Solar Photovoltaic PoweredWaste Plastic Recyclebot System"
https://www.e-helvetica.nb.admin.ch/api/download/urn%3Anbn%3...
That article cites this article from 2006 as its source for energy intensity of solar manufacturing:
"Embodied energy analysis of photovoltaic (PV) system based on macro- and micro-level"
https://sci-hub.tw/10.1016/j.enpol.2005.06.018
That publication finds that silicon purification and processing accounts for the lion's share of embodied energy in solar PV.
But if you read section 6 of the paper, "Embodied energy of silicon purification and processing", you see that those authors are using material production energy intensity numbers from 2004 and 1998. They are also assuming the use of electronic grade silicon for solar manufacturing, and a silicon requirement of 12 grams per watt-peak of solar module. Cheaper and less energy intensive solar grade silicon has entirely replaced electronic grade silicon in PV since the early 2000s. Modern solar module silicon use is about 3 grams per watt-peak, not 12; see Table 1 in https://pubs.rsc.org/en/content/articlehtml/2020/ee/c9ee0245....
What first appears to be a reasonably recent citation for PV embodied energy is actually a chain of painfully outdated assumptions going all the way back to the 1990s.
Also keep in mind that all the panels we tested are much smaller than the ones oin those studies. This means that things like the frame, wires, connections become more important for embodied energy.
I would love to see a life cycle assessment using wholly up-to-date numbers. I keep reading new studies on PV LCA, energy return on investment, and/or energy payback time. People who write these sorts of papers don't seem to keep up with what industry is actually doing. You can learn a lot from data sheets and trade publications. E.g. from published glass thickness and module size and efficiency, you can calculate the quantity of glass currently needed per watt-peak. It's significantly lower than any of these studies using decade+ outdated numbers.
I think part of the problem is one of incentives. Academics writing about LCA are often comparing some hoped-to-be-up-and-coming technology against the mainstream. Like thin film PV, organic PV, or dye sensitized cells pitted against crystalline silicon PV. In that case using old numbers for silicon PV helps the newer technology look like it offers exciting improvements.
Another problem is that reviewers apparently don't care very much about these temporal effects. They don't chase the citation chains to find the really outdated measurements cited in recently submitted manuscripts.
Another problem is that the solar industry has grown large and competitive. Cutting-edge numbers about energy consumption for silicon refinement are probably retained as a competitive advantage by the biggest producers, for example.
It's possible to set tighter upper bounds on resource intensity just from teardowns of recently manufactured modules. I suppose that teardown based analysis may itself be the sort of information you only get from specialty publications like the Photovoltaics International magazine, which is expensive and not indexed by DOI or part of ordinary academic libraries. (So it's not even in sci-hub.)
It's $599 a year if you want to be able to read back issues of Photovoltaics International from their archives:
https://store.pv-tech.org/photovoltaics-international/
I am interested enough in photovoltaic technology that I have bought a couple of $100+ specialty books from academic publishers, but $599 is a bit too steep even for me.
A little bit of knowledge can be a dangerous thing.
At least we're not making many safety critical design decisions based off of them, yet.
Ah, maybe a lot of media content would be seen right away of no interest if those standards had been followed. Or, about news writing, in some old movies,
"If it's not good, I'll make it good."
and
"A good reporter doesn't get great stories. A good reporter makes them great."
Now from that and not being born recently, I have learned to avoid nearly all news media.
The antipode of Barcelona (where this is based) is pretty close to New Zealand.
If they put a second server there and then used a anycast IP, chances are one of the servers would be up at all times with no battery at all.
Edit: Changed multicast to anycast because for some reason my computer wants to auto-correct it. :(
I really wonder how helpful such projects are. Making the Internet greener is undoubtedly an important goal, but I feel this is perpetuating a myth that we're gonna fix the climate crisis with small-scale projects from below.
Practically this is doing nothing to provide any relevant fix for the problem. What we should be doing is thinking about how we can fix the problem at scale, e.g. pressuring large IT companies to get real about the green image they like to peddle. (i.e. care more about news like this https://news.ycombinator.com/item?id=22167858 )
https://sustainability.fb.com/sustainability-in-numbers/
https://aws.amazon.com/about-aws/sustainability/
Unless facebook, AWS, google and Microsoft start encouraging more efficient webiste designs (static websites, limited library usage, no more 1MB fonts, etc.), the work done by the author is absolutely relevant. Limited hardware capacity has always been the most reliable way to limit energy usage.
The battery should be replaced about every 5 years, the solar panels 25 years, the controller every 10 years.
The first time I saw it (I don't have the numbers handy now) I ran some experiments and it seemed clear to me that a JPEG would work much better, and if dithered PNGs were really a good option, more people would be doing them. This was on photographs, where JPEGs are kind of a home-run and PNGs aren't good no matter what you do to them.
This time they're doing diagrams, which would probably be best as regular PNGs - The dithering requires you compress a pattern that's almost noise, and a JPEG would add artifacts without being any smaller.
Here's some other thoughts:
- WebP does exist, but of course you have to do some negotiation to avoid blank images on browsers that won't decode it.
- The site is behind CloudFlare anyway, so if it's a static site with no auth you can probably just put the whole thing on CF / AWS / whatever and it won't use more energy in the cloud than proxying for your own server already does.
- CloudFlare probably has a button that re-compresses everything as WebP for you.
- Economies of scale always apply.
On scale: The transmission losses for the whole US grid is well under 10%. If solar is such a great idea, build a solar farm and run 1,000,000 websites. Or 1,000 houses. It'll be more efficient than putting panels on individual houses or servers. There is no power source that gets more efficient when you have a bunch of individuals running it instead of a power company. Whether the power company is trustworthy is a question of politics, not technology.
This always gets to me when I see EV chargers with VAWTs at a grocery store. If VAWTS are so great, why isn't the grid building them? The grid already has the big wind turbines which are presumably more efficient than a VAWT. So why not buy power from the grid? Because it's a PR stunt.
In short, I wish they'd be more clear about it being a cool thing and not a practical thing. Solar is practical. Wind is practical. At scale.
On the whole it works. Excess PV panel capacity charges battery, ensuring enough backup to run during unfavorable angle, cloud cover, weather, shadows, and night.
Most common issue is re-positioning panels every few hours to favorable angles & avoiding shadows.
Greatest concern is prolonged cloud cover, depleting batteries after a couple days of insufficient light. The cost of preparing backup against “multiple standard deviations” is substantial, buying rarely used batteries (and extra panels to charge them in reasonable time) - hundreds of $ of gear (2-4x base cost) used maybe one day a month. Winter makes this outlier the norm, magnified by its own outliers.
Also, one becomes very aware of app power consumption. Found one web page (AgileCraft logout page) pulls 30 ways for no good reason.
I’m sure solar powered web server would face comparable issues. Depleted batteries are a brick wall, waiting for not just light & time to recharge, but to run the system ASAP.
(it's on the solar powered website itself)
I have seen statistics around a recycling rate of 99% or higher in the US, but also that much of that is done in Mexico or other places with very weak environmental and occupational regulations.
In any case, lead poisoning is a very serious problem which is exacerbated by lead recycling and production.
Lithium batteries can explode when the lithium comes in contact for water. I'm not clear on what the waste stream is doing for all the non-cobalt chemistries (mostly manganese).
I would love to get into distributed web tech. I'm not sure how much of a market there is for it though.
The benefit of being able to have these scuttlebutt networks of low-power, efficient devices is a lower-overall carbon footprint for the common case of serving low-fidelity content like web pages and small applications. As well as the network and content being resilient to local changes in climate events (flash floods, fires, etc). And possibly bringing access to more areas where network connectivity is slow, expensive and unreliable.
Please do this!! I want to see that article on CAES actually worked out in real life not just theory with no howto steps .
https://news.ycombinator.com/item?id=19442938 https://news.ycombinator.com/item?id=19782760
Also online.. https://www.lowtechmagazine.com/2018/05/history-and-future-o...
Down the line, problems are efficiency and all the downside of working with gases.
I don't get it. The carbon intensity of the national grid should result from a life-cycle analysis, so all emissions should be included in the figure. As far as I know, apples and apples are compared, and the home-made version is worse.
They seem to be running on a Raspberry Pi that uses two watts, so they can run Linux. But a website wouldn't have to run Linux. Contiki includes a webserver and can run on an STM32F103. (I'm not sure if the Contiki webserver fits on an STM32F103, though; Contiki is pretty customizable.) They say they have 865,000 yearly visitors, but unfortunately don't explain how many hits that is; if we assume it's 1000 hits per visitor, that's 865 million hits a year, which is 27 hits a second, in the ballpark of what you could do on a 486. So it ought to be within the capacity of a 72MHz 32-bit STM32F103, which uses 50 mA going full tilt — 165 mW if you're running on 3.3 volts. That's better than an order of magnitude less power.
This is probably an interesting experiment to do for resiliency purposes, but I don't think it makes a lot of sense for reducing resource usage in this case. If we assume "Kris De Decker" is the name of a human body that dedicates most of its time to writing this magazine, well, that body dissipates about 100 watts. You could run the magazine on 102 watts by using a 2-watt webserver, or 100.17 watts by using a 165 milliwatt webserver. But if they eat beef once a week, well, beef wastes about 96% of its energy input, converting it to cow poop instead of food; that's 4.8 watts of beef produced from 119 watts of soybeans and corn. By replacing one of those beef meals per year with a vegetarian meal — eating beef 51 times a year instead of 52 — they could reduce their energy consumption by more than the entire web server power budget.
Or, to look at it another way, eating beef once a year uses as much power as the web server: 72 MJ/year, 2.3 W.
(I'm ignoring the embodied-energy calculation because the article shows that it's small compared to the ongoing power use.)
Average marketed energy consumption in the rich world is about 10 kilowatts per person, although typically that figure doesn't include things like corn and beef. Interestingly, in another article https://www.lowtechmagazine.com/2016/05/how-to-go-off-grid-i... the author explains that their laptop uses 20 watts of power, and their external monitor uses 16.5 watts, together 18 times the power used by the web server. If they could manage to do their writing with a USB keyboard plugged into an Android cellphone with an OTG cable, they could probably reduce that to 3 watts, a reduction of 11 times the web server's entire power (although maybe they only write 8 hours a day, so maybe it's only 4 times.) If they could use an incrementally updated e-ink screen, an option I explored in some detail in Dercuano, they could use another order of magnitude less still.
I feel like sustainability is a bigger question than resource use, though. I can't sustain the laptop I'm writing this on because it contains parts I don't know how to fix, even if I could supply it all the energy it needs with like a bicycle generator or something. In fact, nobody in my country knows how to build a laptop like this; a lot of the knowhow only exists in China, and other parts only exist in Korea. Exploiting its CPU backdoors requires knowledge that is presumably only available in certain companies in the US. These seem like much bigger sustainability concerns to me than the really quite minimal power usage of the machine, which is a tiny fraction of the power usage of, for example, a candle (≈80 watts).
For a website that proud itself on being against trends, I would had more appreciation if they went against the trend to use 3rd party sites and be totally on their own. My NoScript reports for them these as 3rd party scripts: google-analytics.com, google.com, googlesyndication.com, gstatic.com, jquery.com, s3.amazonaws.com, statcounter.com, typepad.com
This reminds me of that joke with electric cars that recharge their battery using a diesel generator.
Anyone else seeing this? Looks like they allow me to access it via https://solar.lowtechmagazine.com/2020/01/how-sustainable-is..., but if they actually banned my IP specifically I don't really want to violate their wishes, I just wonder why I'm banned :/
I had a quick look and didn't find an obvious resource online to study. I messaged a couple of popular web hosts and each said they don't have any specific policies on sustainability or any energy usage stats to share.
Not quite how this article meant solar-powered, but still some useful food for thought.
Did they calculate the energy required to construct a human that is capable of powering this server?
In fact that's pretty much exactly what they did. 168Wh battery pack is a small Deep Cycle SLA. A 50w Solar Panel and associated charge controllers and the like is like $80 at Harbor Freight. The whole thing is quite achievable on a budget.
I think you mean anycast, not multicast, but a less exotic option would be to use DNS failover, or even just round-robin DNS with no explicit failover: https://webmasters.stackexchange.com/questions/10927/using-m... https://www.nber.org/sys-admin/dns-failover.html
Nuclear usually needs to keep running, and fewer people use electricity at night (unless you live in Philippines where it is common to only run AC at night to save on electricity).
A battery system is almost certainly easier than physical servers spread across the globe, even when you account for the cost of the battery.
If you're going down that route anyway, you could add redundant servers at different latitudes to hedge against cloudy weather.
Only one of them needs to be up at a time. By choosing the antipode, by definition one of them will be in sun when the other is not (weather notwithstanding). The equinox would be the hardest day to deal with because they would both be at low energy at sunrise/sunset.
So yes, you're right, a third server would probably make it work almost 100% of the time.
Problem solved.
Large-scale projects don't magically wink into existence in a vacuum, the conversation needs to happen and the ball first needs to get rolling.
The nice thing about that is that you don't have to change the tech much, you can still use digital ocean or whatever, but they need to get their power from renewable, which in turn means their grid needs to.
The not so nice thing is we are not changing fast enough. I hope pure price pressure from tech advantages will get us from fossils to renewable.
at least this shit is inspiring
This is pure speculation on my part, but probably much less efficient. In order to make the p2p network reliable, you'd need many more copies floating around. I also suspect that "finding" your data is more energy intense compared to basic DNS lookups.
It's also possible to cache lookups in the same way as DNS by having larger nodes cache the lookup data and having smaller nodes query them, so that the most common queries are satisfied using a single request to a lookup cache.
Meanwhile the advantage of a P2P network is that most of the nodes are client devices which would have been powered on regardless, instead of needing additional devices dedicated only to hosting data.
Of course, there's always IPFS, but that project comes with its own issues (e.g. modifying content).
What is the marginal cost of the web traffic it creates and replaces? Would tuning the software and data payloads be more impactful than worrying about the server wattage?
What is the marginal cost of other user activities which it influences? Not just the website operator, but the user behaviors happening as a result of their relationship with the service? Do they stop using other less efficient services or just increase their overall footprint? Could it reduce their consumption of energy and material goods? Change their diet or travel habits...?
The lead acid battery industry is incredibly competitive and prices unbelievably low due to this vertical integration and mechanized production. Pretty much the only way to beat the U.S. car battery prices is to subsidize (Korea) or pollute like hell (Mexico).
The U.S. is a net importer of lead, largely because all of the SLA batteries coming in from Asia end up in the waste stream as scrap, so there isn't very high demand for virgin (smelted lead concentrate) when all that scrap is sitting around.
Hydro, nuclear, wind turbines would all be 0g/kWh if it was only about fuel.
I'd like to see what panels those are, because the ones I've built while working as a PV manufacturing tech, both mono and poly (roughly 21% efficiency,) will have greatly varying voltages with even the tiniest hint of cloud cover over one cell, even with the junction box working to help separate out sections of the panel to maintain better voltages. Typical 60 cell 30-32V panel will drop to ~18-20 with just two cells on one 20-cell section of the panel covered. Sure this is still enough for the paltry voltage this specific server needs, but if they used smaller and more affordable panels like those used for cell phone chargers or similar size (within about 18"x18" form factor,) I can guarantee you those do not take to shading or even bad orientation well at all. 45 degrees off direct-exposure and you could be looking at that smaller panel producing a mere 2V or less.
I still wouldn't want to run a Pi directly off solar. I'd want enough warning to shut it down rather than killing the power.
As a practical matter insolation at dawn/dusk won't be able to power much, without a PV array that would be quite oversize during the day.
Lots of interesting optimisation problems in this area. But at this scale batteries and solar panels come in discrete sizes, so it's a bit academic.
Bad assumption: Most people use laptops, which aren't "powered on regardless", or mobile devices, where the power consumption of being active on a P2P network will kill the battery.
Moreover, the average mobile device spends something like eight hours a day (while the owner is asleep) connected to the charger where battery life is irrelevant. So give the network two modes, one where the device participates fully in the network and is a preferred source for data and one where it only registers what it has with another node and is then a provider of last resort for data not otherwise available elsewhere, and operate in one mode or the other depending on whether the device is running on battery.
Notice also that the amortized number of hits over the whole network is one per object per device, i.e. the average number of times a node has to upload each thing it downloads is only once. That doesn't take an insurmountable amount of battery even if you couldn't shift most of the load to the devices that are plugged in.
The network might be slower overall but we're talking about optimizing for lower power consumption so I feel like that would be a lesser priority.
Now I'm starting to wonder if there is a away to simulate the power draw of such a network.
For the solar powered website article, it's the order of magnitude that matters. You say I overestimate the energy use of solar panel production, but in our configuration it corresponds to just 1 liter of oil per year.
If the numbers you were using were correct, then just the energy input for the solar panels would have cost more than the wholesale price for the modules.
For example, How do you account for the fact that all production facilities for solar panels have moved to China? If you look at the price evolution of solar panels, there's a gradual decrease due to technological progress (less energy use indeed). Then, from 2009 onwards, the decline in costs accelerates sharply, the consequence of moving almost the entire PV manufacturing industry from western countries to Asian countries, where labor and energy are cheaper and where environmental restrictions are more loose.
https://solar.lowtechmagazine.com/2015/04/how-sustainable-is...
Depending on incentives and purpose one can choose which facts that one want to use. A fair comparison would be quite complex and have many variables one need to define for both the solar version and the power grid version.
The other approach compare the state of the art PV module vs the state of the art power plant, both being built using the latest mining and construction methods.
The answer of what is more sustainable is going to depend a lot on context. Are we comparing what a person can do right now, or what we should be doing in the future? For powering websites the general answer from the industry is currently to build data centers in depopulated areas in countries with cold climates, near a hydro power plant and with plenty access to cooling water. This is however not a solution that you can apply if the requirement is that the server must be located in Spain.
In my experience running large websites, that's about 10% of the internet, if not more.
When I made a DNS change, only about 70% of the traffic dropped off in the TTL. The rest took anywhere between a few hours and a few weeks (and some never dropped off, we had to just let them fail after a while).
Do you think the advent of DoH will improve that situation?
We test by shading different cells at a known light level, and seeing how far the panel as a whole deviates from our expected baseline. Usually what we'll find is that a cell hadn't had its ribbon bars properly soldered while in the tab/string machine, so the shading of the cell dramatically reduces performance. You can't catch this kind of defect easily in an EL tester, as usually you've got enough contact between ribbon and cell to get the entirety of the cell to emit light, thus hiding defects from soldering. So we implemented that shading test to root out bad cells/bad soldering jobs. Quite proud of having thought of that testing methodology, because I've not seen any other company test panels in that manner.
But the costs of solar panels made in China in 2020 are much lower than those of solar panels made in China in 2010. Chinese factory wages have not fallen over the past 10 years, nor have Chinese pollution controls been further relaxed. The trend of falling costs over the last 10 years of made-in-China solar panels is due to improved techniques of production.
In 2010, Chinese producer Suntech was the largest PV manufacturer in the world. It sold 1.6 GW of panels and realized revenue of $2.9 billion:
https://www.greentechmedia.com/articles/read/suntech-q4-earn...
Its panels were therefore selling for about $1.83 per watt in 2010.
Presently, Chinese high efficiency monocrystalline PERC panels are selling for no more than $0.25 per watt:
This 85% drop in prices indicates that technological and manufacturing progress in the solar sector continued at a rapid clip even after its center of gravity shifted to China.
You take the solar panel that just comes out of the factory as a reference, but I could just as well argue that the reference should be the age of the average solar panel installed in the world.
Also, let's wait and see how the move to China affects the quality and life expectancy of solar panels. And what it means for other forms of pollution. I focused on energy use but there are other environmental concerns with the production of solar panels.
I will be happy to update the calculations in the future if more and better data are available.
You'll still be talking to your local DNS server with its own caching rules.
Chrome for example will only use it against your ISPs DNS servers, otherwise it will use Google's servers. So that would at least help, but if the big ISPs start supporting it, then Chrome would use that.