A question about the methodology - how is the carbon measured?

[mrchrisadams]

Hi there,

I'm not a salesforce developer, so it might be really obvious to people who are (apologies if so), but is there a chance of expanding on the methodology below to make clearer where the carbon intensity figures for the electricity come from?

I see this in the methodology section of the README:

Carbon per mms of Core App Server time is derived from a PUE-adjusted carbon coefficient per data center for both AWS and Salesforce first party data centers. This figure is calculated as the emissions factor in grams per megawatt hour multiplied by the PUE per datacenter. The PUE-adjusted carbon coefficient is then scaled against each App Server SKU types’ daily megawatt hour and divided by the available processing seconds, in millions, for that specific App Server SKU to get to the estimated carbon per million milliseconds of compute time. These per-SKU, per-data center carbon coefficients are then combined with data on the App Server SKU types that serve each Salesforce instance to produce a single carbon coefficient per instance. Data on the App Server SKU types comprising Salesforce instances are current as of current as of June 2023.

There's a bit specifically here:

This figure is calculated as the emissions factor in grams per megawatt hour multiplied by the PUE per datacenter.

Is this based on average carbon intensity per region? And is it location based or market based? I'm asking so I can understand how you might integrate figures from use of salesforce as part of a portfolio of software products used by a company.

[tysonread-salesforce]

Hey Chris, we used regional carbon intensity. One caveat to add here is that this dashboard only reflects the usage of a handful of features in Salesforce, it doesn't show you a footprint for all of Salesforce. It's really intended to help admins/developers use those tools in a more sustainable way.

On Mon, Oct 23, 2023 at 3:37 AM Chris Adams @.***> wrote:

Hi there,

I'm not a salesforce developer, so it might be really obvious to people who are (apologies if so), but is there a chance of expanding on the methodology below to make clearer where the carbon intensity figures for the electricity come from?

I see this in the methodology section of the README:

Carbon per mms of Core App Server time is derived from a PUE-adjusted carbon coefficient per data center for both AWS and Salesforce first party data centers. This figure is calculated as the emissions factor in grams per megawatt hour multiplied by the PUE per datacenter. The PUE-adjusted carbon coefficient is then scaled against each App Server SKU types’ daily megawatt hour and divided by the available processing seconds, in millions, for that specific App Server SKU to get to the estimated carbon per million milliseconds of compute time. These per-SKU, per-data center carbon coefficients are then combined with data on the App Server SKU types that serve each Salesforce instance to produce a single carbon coefficient per instance. Data on the App Server SKU types comprising Salesforce instances are current as of current as of June 2023.

There's a bit specifically here:

This figure is calculated as the emissions factor in grams per megawatt hour multiplied by the PUE per datacenter.

Is this based on average carbon intensity per region? And is it location based or market based? I'm asking so I can understand how you might integrate figures from use of salesforce as part of a portfolio of software products used by a company.

— Reply to this email directly, view it on GitHub https://github.com/seamusocionnaigh/DeveloperCarbonDashboard/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/ARUWOVYP5ITHQXCNALBATI3YAZCHNAVCNFSM6AAAAAA6L3FTIOVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE2TMOBWGMZTEMY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[mrchrisadams]

Thanks Tyson - when you say regional carbon intensity, would you let me know what resolution, and is this average intensity, or marginal intensity?

I ask as marginal and average can incentivise different interventions:

https://www.electricitymaps.com/blog/marginal-vs-average-real-time-decision-making

For example Electricity Maps provides data at a balancing authority level (basically the grid, not the nation state, or even the state) https://www.electricitymaps.com/data-portal

I found out this year that this is published at an hourly resolution for consumption, with a few days lag. The repo has some python code for fetching this data from government sources (the IEA in the states):

https://github.com/mrchrisadams/hourly-carbon-intensity-usa