Has anyone here come across a good methodology for working CO2 footprints of virtual machines?

Has anyone here seen a good bottom-up methodology for working out the CO2
footprint of say, a running server that's provisioned by the hour, like
cloud servers are?

My current mental model is basically what's below:

   1. take the energy used by a whole physical server
   2. divide by the number of virtual machines it is probably is 'sliced'
   into
   3. divide for a single vm, the percentage it's in use (i.e. how many of
   the 740 hrs of a month, it's 'on' for)
   4. multiply by the PUE
   5. multiply by CO2 intensity of power used (nukes are low, coal is high)

This should give a single CO2 figure, that might work on a per month basis.

This doesn't take into account the life-cycle-impact making the *server*,
but you can amortise this over the expected lifetime of a machine.

It also doesn't take into account how machines in a larger DC are switched
on or off in response either, nor changes related to serverless or
containerisation of applications.

I'd be curious to see if anyone on this list has been working on ways to
quickly and easilty calculate this - if you folks can point me to any good
papers covering this I'd be really, really grateful.

Cheers,

Chris




-- 
See when I'm available for a call: https://calendly.com/mrchrisadams
Subscribe to my newsletter about digital product development:
http://bit.ly/prod-sci-method

Chris Adams
email: chris@productscience.co.uk
www: productscience.co.uk
skype: chris.d.adams
tel: +44 203 322 5777
twitter: mrchrisadams
mob (UK) : +44 7974 368 229
mob (DE) : +49 1578 474 4792