What's the carbon footprint of stackoverflow?

Question

Inspired from What’s the carbon footprint of an email? and other recent questions that made me reconsider the "immaterial" internet usage.

I am curious about updated statistics about the carbon footprint browsing the internet (it should probably be divided by service, since facebook and gmail recently claimed to be have moved to sustainable data centers, while I haven't any idea about amazon web service, or stackoverflow :) ) and about computer usage in general.

I found a nice idea in cloud carbon, a chrome extension that shows your carbon footprint, but it's still far from accurate (for example, looking in the source code, you see that they just estimated facebook carbon footprint equal to the computer usage). There is also a project from Microsoft but I can't try it now.

Answer 1

Not a full answer to your question, I do not know the footprint of the servers running the StackExchange sites, but it could be estimated from the information here. Good performance automatically means it uses as little resources (energy included) as possible.

You could measure your own carbon footprint while using StackOverflow if you are using an UPS that also measures your use. See my answer to the question "Is it possible/easy to determine how much power a program is using?".

Here's a screenshot of what my computer is was using after a day of rendering 3D (Povray) scenes:

The figures may seen shocking, but rendering 3D scenes uses 100% processing capacity of all 8 cores in my workstation. Browsing any site, including Stack-Exchange sites, would only use a fraction of that. Unfortunately I cannot do a measurement of only using Stack-Exchange at this moment. To do that I would have to stop the current process which would be the same as throwing away a few hours worth of rendering (quite some CO2) which I am not prepared to do :-).

On a laptop or tablet, the footprint will be allot less than on a desktop or workstation, so a real estimation could only be made by knowing what type of devices are connecting (server logs could provide a good estimation I guess).

Answer 2

Here's a back-of-the-envelope calculation:

This page describes the architecture of the Stack Exchange network. It consists of:

• 9 web servers
• 4 sql servers
• 2 redis servers
• 3 tag servers
• 3 elasticsearch servers
• 2 haproxy servers

So that's 23 servers in total. Not all servers are equal, but since we do not have exact energy consumption figures we will use an estimate for 1 web server and multiply by 23.

Recent measurements done by an organization I cooperate with showed that their server energy consumption is between 1160 and 1770 kWh a year. Let's take 1600 kWh as a rough estimate of server energy consumption per year.

The EPA reports 7.07 × 10-4 metric tons CO2/kWh as the average carbon footprint for 1 kWh in the US, but that average kWh is generated with only 17.64% renewable energy. This meta post says that the Stack Exchange servers are hosted in New Jersey by a company called QTS. In the QTS Sustainability Report 2018 it says that 32% of their power is from renewable energy sources. It's not entirely correct, but we estimate that the carbon footprint of 1 kWh used by a QTS server is 0.707 / (100 - 17.64) * (100 - 32) = 0.584 kg CO2e per kWh. So for 1 server this results in 1600 * 0.584 = 934 kg CO2e per year.

So far we only looked at the carbon footprint of the 'use phase' and ignored the production and disposal phase. This 2010 paper from Dell says that the GHG emissions in the use phase of a Dell rack server accounts for over 90% of the total life-cycle impact, so we add another 10% to get the impact over the entire lifecycle of the server: 934 / 90 * 100 = 1038 kg CO2e per server per year

With 23 servers that results in 23,900 kg CO2e per year for the entire Stack Exchange production environment (so not including servers for developing or testing the platform).

With 1.3 billion page views per month that is approximately 1.5 mg CO2 per view