Ask HN: How can I calculate CO2 emission of servers/VPS?

Have you contacted Hetzner? They claim 0% emission impact https://www.hetzner.com/unternehmen/umweltschutz/

Because your server is not emitting CO2, whatever you make up, is just something you made up (best industry practices, conversely, are fictions you choose to believe).

If you really care, put a watt meter on it and try to reduce its power consumption.

Because you don't have control of the data center's grid connections.

Good luck.

With AWS last I checked (which might have been 2 years ago) they advertised regions powered by renewable energy sources. You can see it better here: https://sustainability.aboutamazon.com/around-the-globe?ener...

But as to calculate exactly the amount of carbon footprint for your VPS, I think you are out of luck, because exact numbers are either opaque or required you take AWS at their word.

Are you looking for total carbon on carbon per hour?

Because the second can be estimated via electrical use. But the first involves the building of the machines etc.

https://engineering.teads.com/sustainability/carbon-footprin...

https://docs.aws.amazon.com/awsaccountbilling/latest/aboutv2...

My guess is these will get within an order of magnitude.

I used to work at an research/architecture (buildings, not software) firm that was a leader in carbon emission estimation. To estimate the carbon emissions from building operational energy, what we did was find the rate of carbon emission at the local grid[1] and then multiply it by the predicted operational energy from an energy model. For example, for a single family home drawing from a blended grid mix of coal, gas, and nuclear, you might have a carbon emission rate of 150 kgC02/MBtu for 0.04 kBtu/ft2 of operational energy, over the course of an hour. This works out to:

    150 kgC02/MBtu * 4E-5 MBtu/ft2 = 0.006 kgC02/ft2

You then integrate this over the course of the year. This gets complicated as there are variations in the resolution of carbon emissions from different grids, the grid emission rates, and which grid the building uses.

To estimate server carbon emission rate, you just have to swap the building energy use intensity for server energy use, and everything else should be the same.

[1] Grid Emission Rate Sources EPA: https://www.epa.gov/egrid/data-explorer WattTime: https://www.watttime.org/

It all depends on what the energy mix of the grid supplying the data center looks like. This outfit seems to have a tool that keeps up to date on that:

https://www.datacenterdynamics.com/en/news/cirrus-nexus-laun...

> "The tool takes in data on carbon intensity from local utility grids, and relates that with the reported strategies of cloud providers. It also calculates energy usage based on the manufacturers' data for the servers involved and PUE from the data center operators."

See also:

https://techxplore.com/news/2022-08-centers-fueling-climate-...

Getting accurate measurements (i.e. say +/- 10%) looks rather difficult, as regional grids vary their energy sources seasonally and even daily, due to things like wind/solar variability, hydropower output, coal vs. gas input, etc.

There are various scopes to consider.

Scope 1: the fuels you burn on site to generate heat/cooling or electricity.

Scope 2: the electricity you import on site probably burns carbon so it has a carbon factor you can find that on your utilities website (most likely). Imported electricity in kWh x carbon factor tCO2e/(delivered)kWh = emissions in tCO2e - you may want to start here with the location, grid electricity and expected electricity use.

Scope 3: are the emissions made off site to make materials that are imported on site to make the business run.

There are no hard rules to do this analysis. Most of it is a linear approximation. Most people don’t publish analysis beyond scope 1/2.

It’s really important to list assumptions and sources. Good luck.

ThoughtWorks made an app that does this: https://github.com/cloud-carbon-footprint/cloud-carbon-footp... but mainly works with aws, azure, and gcp

You might be interested in their methodology page https://cloudcarbonfootprint.org/docs/methodology/

Full disclosure: My company was one of the early testers of the app

Autumn8.ai (www.autumn8.ai) provides accurate benchmarks of AI models for energy consumption and carbon emissions, both in the cloud and for on-premise servers, prior to deployment. Further, they also offer performance metrics (latency, throughput and costs) with no server spin-ups. This helps customers choose the optimal server to reduce CO2 emissions while meeting other SLA requirements. Please contact Vanya Amla (vanya@autumn8.ai) for more details.

We had a hard time finding this data when experimenting with carbon aware scheduling in Nomad. There seem to be a lot of orgs working on this, but it’s all third party as far as I know. I couldn’t find a first party (eg Amazon, Microsoft, Google, etc) that provided this data via an API. Very frustrating.

There are a couple resources linked here: https://github.com/hashicorp/nomad/blob/h-carbon-meta/CARBON...

The silver lining is that the folks who are working in this space are extremely friendly and passionate in my experience. Don’t be scared to “contact sales for an API key.” The few folks I talked to were extremely helpful.

In the most recent episode of the Last Week in AWS podcast Corey Quinn just talked about this topic https://www.lastweekinaws.com/blog/how-google-cloud-and-aws-...

For AWS, there is also Customer Carbon Footprint tool available https://aws.amazon.com/about-aws/whats-new/2022/03/aws-launc...

Google provides first party data for GCP https://cloud.google.com/carbon-footprint

as well as icons that nudge you if you work through the console

I recommend you take a look at co2.js. This was just mentioned in the latest issue of branch magazine (https://branch.climateaction.tech/issues/issue-4/co2js/) and I haven't played with it myself yet. Code is linked from the article but also https://github.com/thegreenwebfoundation/co2.js/

Does running Data Center workloads on renewable energy is any selling point to some corporations? Or it is a too minor feature? I happen to work at one of those places, but I am thinking it is just a marketing/PR value add-on: https://www.deac.eu/news/information-for-the-media/data-cent...

Have a look at the Boavizta API. It features an automated evaluation of environmental impacts of ICT services and equipments.

That seems to cover on-prem servers and AWS.

https://boavizta.org/en/blog/boavizta-api-automated-evaluati...

https://github.com/boavizta/boaviztapi

You cannot since you will never get access to the required data to make such a calculation. All you can do is take the company that provides the services for you at their word.

I doubt you'll ever get an exact measurement from big companies like Amazon. If you can know your exact CO2 emissions for a workload you can pretty accurately model the power consumption of the processor. If you're Amazon and running entirely custom silicon like their Graviton processors then that power consumption info is likely a closely guarded trade secret for now.

Very tricky. You need at least:

- the electricity consumption of the system, which varies depending on use

- the electricity consumption of the air conditioning, which is to a first approximation about the same as the fist number (you've got to remove all that energy again, at some efficiency coefficient). Varies by ambient temperature.

- the CO2 mix of the grid at that time, which varies from minute to minute.

Maybe helpful:

https://github.com/mlco2/codecarbon

Also:

https://github.com/hubblo-org/scaphandre

I built a tool for benchmarking CO2 emissions for Go programs once. It was done as a joke for a meetup once, but is now more relevant than ever.

https://github.com/zegl/co2go

Try https://www.websitecarbon.com/website/news.ycombinator.com/ and just substitute the desired domain name

Check these peeps: https://www.climatiq.io/ They also have a case study specifically about servers.

In AWS you can view your specific emissions under billing > cost & usage reports.