Emission factors for use of server space

Hello,I’m working on a project with a customer that offers sofware (webservice) solution to end-customers. For the webservices, that are developped and hosted by third party providers, data is stored on cloud or on-premise servers. Users call upon the webservices several times, for certain hours a day.
I’m looking for the best way to estimate the carbon impact of this server space (cloud and on premise) that the customer is using. What are the emission factors you are using? How do you consolidate the information?

3 Likes

Hello Sven

Very relevant question!
Cloud usage emissions can be very hard to get a grip on, since it is not easy to do an order-of-magnitude estimate without a lot of reference points.

High level overview:
Cloud consumption is all about energy consumption (of the servers).
And server load is function of how computationally intensive it is used and for how long (expressed in something like CPU-hours).
Secondary aspect is just storage use (expressed in GB). Though for most applications this one is a lot less dominant than the first.

In practice: source data

1) Do it yourself (hardcore)
In practice it all comes down to source data.
What do you know yourself, and to what detail of information do you have access?

With the aid of a specialist, you might get to some approximation for the load-hours and the total-storage used.
Additionally, you need to find or deduce the location of the servers and use that to get an electricity emission factor (CO2/kWh)

A lot of data will be hard to find, and there will be some guesswork involved. But this method will get you the most tangible information, and also the one that gives the most insights to improve upon!

2) Ask your cloud provider(s)

Most cloud storage providers have some sort of support for collecting data related to your own cloud carbon footprint. One is a more helpful and detailed than the other though.
You’ll need the help from a tech person at the company itself to access these dashboards.

From best to worst:
A) https://www.cloudcarbonfootprint.org/ (all applications)
If you can get this open source-project running on all relevant accounts and servers, you will get a very clear and easy to use dashboard of emissions

B) Azure (Microsoft)
See https://www.microsoft.com/en-nz/sustainability/emissions-impact-dashboard

C) Google Cloud
See https://cloud.google.com/carbon-footprint

D) AWS (Amazon)
See Carbon Footprint Reporting – Customer Carbon Footprint Tool – Amazon Web Services

For all of these the following is true:

Try to read out the carbon data as “high up in the chain” as possible.
If you can get kWh of electricity data, you can take that as your input.

Copying their total CO2e-factor is less ideal, as you’d prefer to have knowledge about the assumptions behind that number.

3) Last resort option
If none of the above work well, you can use the total amount paid for cloud services as an upper bound for your cloud carbon emmissions.

If you pay 100.000 EUR in cloud services, that can not be consuming more than 100.000 EUR in electricity costs.
(taking average retailer prices in the region Electricity price statistics - Statistics Explained)
(Otherwise they would be losing money on that service)

This gives you an upper bound, and you can apply an additional scaling factor (% of service price is energy cost) that feels reasonable to you, to get a rough first estimate.

I hope this was helpful!
Felix

4 Likes

Thanks Felix for the information, very insightfull. Espeially the Cloud Carbon Calculator is an interesting tool to investigate.

1 Like

Thank you for your response, Sven!
Happy to help.
And I agree that Carboncloud is a wonderful tool!

I just want to add that the pre last resort would be to utilise Exiobase as a web hosting service [72.30.23] falls under the category of Computer and Related Services (product)

3 Likes