Very cool. I’ve noticed that when things aren’t measured, people tend to be wasteful. At a previous org, Aws costs weren’t made visible to the developers and spending was just going through the roof. To fix it, we set up a weekly tracker that would notify you of your spending and include a breakdown by service.
Today it’s very opaque to me how bad the carbon emissions are for tech, but also it’s hard for me to know what is a good amount in relation to the Paris agreement. So I’d love to have tools to better measure and understand the company carbon emissions for tech. Otherwise it’s also hard to gauge if what people are doing actually matter, or if it’s just theater.
I think carbon emissions are a bad metric, and I prefer the simpler metric of energy consumption for the simple reason that the latter is easier to quantify.
Carbon emission and energy consumption are only officially highly correlated when electricity is mainly produced via fossil fuels. Renewable energy sources however also have a net cost, e.g. windmills need 12-14 years of operation to break even the energy consumption of their production. These costs are often overlooked or not factored in, or we even delve into the trickery of carbon certificates which many companies use for greenwashing.
All in all, the carbon emission metric thus factors in two separate things: Energy efficiency and energy markup. Our code can never change the latter, only at most factor it in (e.g. as is in green HPC where jobs are scheduled depending on the ratio of renewables versus non-renewables), and the latter can even be used to falsify data.
To give an extreme example, which code is better: code A that requires 1 kWh of fossil energy to solve problem X or code B that requires 2 kWh of renewable energy to solve problem X? Surely you would say ‘code A but using renewable energy’, which is where I agree, but at this point you can see how quickly the energy markup argument dissolves, can’t you?
e.g. windmills need 12-14 years of operation to break even the energy consumption of their production.
That seems very unlikely. That’s maybe half the expected lifetime of a wind turbine, and the energy cost of production is a small fraction of the total cost of production, so if it’s true then wind turbines cannot be economically viable.
Some searchengineering says a wind turbine pays for itself financially in 5 - 10 years, and recoups its production energy in 6 - 12 months.
You can absolutely change energy markup, even if the code can’t, by running on different cloud providers for example. Hetzner uses 100% renewable energy. Or you can also set up your own solar powered website like they do at lowtechmagazine.
Renewable energy sources however also have a net cost … the energy consumption of their production. These costs are often overlooked or not factored in, …
Very cool. I’ve noticed that when things aren’t measured, people tend to be wasteful. At a previous org, Aws costs weren’t made visible to the developers and spending was just going through the roof. To fix it, we set up a weekly tracker that would notify you of your spending and include a breakdown by service.
Today it’s very opaque to me how bad the carbon emissions are for tech, but also it’s hard for me to know what is a good amount in relation to the Paris agreement. So I’d love to have tools to better measure and understand the company carbon emissions for tech. Otherwise it’s also hard to gauge if what people are doing actually matter, or if it’s just theater.
There’s also this Peter Drucker quote:
I think carbon emissions are a bad metric, and I prefer the simpler metric of energy consumption for the simple reason that the latter is easier to quantify.
Carbon emission and energy consumption are only officially highly correlated when electricity is mainly produced via fossil fuels. Renewable energy sources however also have a net cost, e.g. windmills need 12-14 years of operation to break even the energy consumption of their production. These costs are often overlooked or not factored in, or we even delve into the trickery of carbon certificates which many companies use for greenwashing.
All in all, the carbon emission metric thus factors in two separate things: Energy efficiency and energy markup. Our code can never change the latter, only at most factor it in (e.g. as is in green HPC where jobs are scheduled depending on the ratio of renewables versus non-renewables), and the latter can even be used to falsify data.
To give an extreme example, which code is better: code A that requires 1 kWh of fossil energy to solve problem X or code B that requires 2 kWh of renewable energy to solve problem X? Surely you would say ‘code A but using renewable energy’, which is where I agree, but at this point you can see how quickly the energy markup argument dissolves, can’t you?
That seems very unlikely. That’s maybe half the expected lifetime of a wind turbine, and the energy cost of production is a small fraction of the total cost of production, so if it’s true then wind turbines cannot be economically viable.
Some searchengineering says a wind turbine pays for itself financially in 5 - 10 years, and recoups its production energy in 6 - 12 months.
Sorry, my data was wrong.It’s 12-14 months and I had it stored wrong in my head. Here is a very detailed listing.
You can absolutely change energy markup, even if the code can’t, by running on different cloud providers for example. Hetzner uses 100% renewable energy. Or you can also set up your own solar powered website like they do at lowtechmagazine.
Relevant Wikipedia entry: Life-cycle assessment § Life cycle energy analysis
Thanks!