Actually solar powered technology infrastructure

1

I recently read the “Actions for Environmental Justice from Autonomous and Community-Based Technological Infrastructures” report and wanted to continue the conversation here. If you are short on time and want to skip to the punchlines, go to page 75 of the PDF.

The study took a broad look at the issues from energy usage, energy generation, e-waste, and similar issues. Because it covered such a broad array of topics, it didn’t go into great detail about any one specific thing.

As an environmentalist, and someone who runs IT infrastructure for my local community, a series of things that stuck out to me were:

  • Evaluate the possibility to set up autonomous, small-scale or locally-run data centers where feasible
  • Explore containerization, virtualization and lightweight options for better common goods use
  • Evaluate and choose data centers based on real sustainability commitments (e.g. location, transparency)
  • Implement local renewable energy systems when feasible, but keep questioning this as a definitive solution, making visible how it can affect communities

This is because I run lightweight virtualization, on my small-scale, locally-run server room. It’s partially powered by solar panels (and would be entirely powered that way but we ran out of room for panels). We also have a battery backup which would keep systems going for at least 24 hours of a power outage (and we confirmed that our internet service does in fact stay up during a grid outage).

So yeah, these ones really hit home for me, quite literally.

An idea I’ve had for a while now is to have a small number of servers, perhaps just one rack that was powered entirely by renewable electricity. This means having batteries to keep things on when there’s no generation (no solar at night, possibly no wind). It also means generating significantly more energy than is needed sometimes (e.g. summer) so it’s possible to generate enough in low times (winter, cloudy, etc.). The same things applies to wind power, as it’s also intermittent.

There would also need to be some type of agreement on how to handle an extended shortage of generation. One option would be to issue a warning and have services need to be migrated to another location (e.g. a traditional data center).

An alternative would be to have the power grid as a backup. If needed, buy energy and perhaps buy carbon credits to atone for using fossil fuels (the majority of virtually all grid power is non-renewable and it’s not like you can tell the electrons from coal power where they need to flow). That could also help with the overabundance of power being generated sometimes, as instead of having that capacity be wasted, it could be sold to the grid and help reduce the need for those fossil fuels. There are trade offs either way.

To provide some concrete numbers, all the servers and networking gear in our rack consumes about 1kW. This means to be 100% renewable, we’d need to generate 24kWh of electricity per day. In July, we can do this even on a rainy day. In December, it is possible, but that only happened 5 days last December. It was also under 5kWh for 5 days in a row. The lowest was 1.8kWh. All of this is with a 17KW solar panel system.

Now, if we could partner with a sister 100% clean datacenter in the opposite hemisphere, that would change everything! Partnerships is also something mentioned in the aforementioned report.

I’m curious as to what fellow members and activists think of this. Two tiny data centers, that were community owned, running efficient IT systems, using hardware as long as possible and then making sure it gets repurposed by others or recycled in the worst case scenario. Little to no reliance on the power grid to keep things going. No water usage to keep machines cool like the big AI data centers reportedly have.

I also have ideas on how this could scale beyond just two data centers and have users not even notice when one of the centers go offline. I also have ideas on how this could be tested incrementally (as opposed to having to buy land, solar panels, inverters batteries, and so forth all up front). I also have ideas on how to help fund this (the short version of which is that there are a bunch of technology lovers who have servers in their basement who might be glad to have them run somewhere like this and pay a fee to cover operational costs of running the hardware). But if I include all the details on these ideas, this post will be as long as that 93 page report. :laughing: So I’ll cut it off there.

What do you like about this? What doesn’t make sense or do you think wouldn’t work and why? If there were a pilot project to show this is not possible to run a sustainable IT infrastructure, do you think it could get funded (crowdfunded as a co-op, funded by grants, etc.)? Could there be enough groups interested in the mission that they’d be willing to pay even though they could get cheaper service from companies like AWS, Microsoft, or Google?

2 Likes
2

I’d really like to see and help make this happen. It’s excellent to be finding this community (in this case, specifically, you) here.

Perhaps one avenue would be community members taking over ownership of existing data centers where they live, and for those data centers that the communities decide not to simply deconstruct or repurpose entirely, perhaps any remaining could be used in this way, too, if that makes sense. Thinking out loud, here though…

3

I’m not aware of any datacenters that are currently up for sale near me, and if they were, I suspect our community would be outbid by the megacorporations.

However, the strategy of making sure permitting requirements are in place and they include showing proof they won’t put an undue burden on the local community’s water and electricity supplies seems like it’s been somewhat effective. So at least the expansion is being slowed down.

I feel like the ultimate form of protest is to just not use what these companies are producing. Let them buy overpriced data centers that don’t make any money.

I still believe that a people’s owned datacenter will look a lot different than the corporate ones. More decentralized. Less capital investment. More re-use. Driven by our own priorities instead of purely cutting costs and increasing prices.

To bring this idea from a pipe dream to a reality will require a coalition of people with different skills, the biggest one being finding some groups who just want a static website and would be willing to work with us as we get established. The next one on my list would be business skills, and optionally finding funding (grants, philanthropy, bringing together co-ops and/or launching a new one, etc.).

For dynamic websites or things that absolutely need extremely good uptime, or want a turnkey solution, I imagine MayFirst hosting would be a better option for the foreseeable future. If it works, maybe it could be a co-location option for MayFirst.

The technology to accomplish this has been around for a long time, and there are multiple tech people I am connected with who would be on board. That’s all completely managable. It’s the business and outreach that is the hard part. There are so many different offerings out there, it’s hard to get noticed.

Anyway, that’s just my 2¢. Hopefully, as time goes on, we’ll keep collecting more people who are interested and willing to put time and effort into this.

4

I would love to see a coop and/or values aligned data center that we could move our servers to. I suspect other providers like May First would also be interested, especially those that are much bigger that May First - like Koumbit or GreenHost and others. Unfortunately, GreenHost at least is geographically a lot farther away. And, it would take a lot of trust and confidence in a data center’s longevity to make the move worth it.

I seem to recall some folks trying to start something in the Bay Area but am having trouble finding any info about it (but I did find this article on coop data centers). Also, I recall someone coming to Aspiration’s Dev Summit and presenting on data centers owned by the state of california (see this link I think - so more of a socialist approach).

Also, at least in this exact moment in time, I’m not sure the mega corporations have any interest in the type of data centers we would need. I took an activist tour of the hyper scale data centers in Loudoun county Virginia last year and, as someone who has spent a lot of time in one of the pre-AI data centers that house the May First servers… wow, these are totally different. The mega corps are competing for data centers that are hundreds of times bigger than the single floor in a 12 story building in New York. They will only buy or build where they can literally have their own electrical substation.

This is way out of my depth of knowledge - but I have no idea what this hyper scale data center expansion means for the kind of data centers we use. How does it affect the economics of the company running our data center? Since the hyper scale data centers are being used for something entirely different than what traditional data centers are used for, is there an impact? And what will the impact be when the AI bubble bursts and we suddenly have massive numbers of hyper scale data centers that are way too expensive to maintain for the uses we would need them for?