Summary
This section provides an overview of the methodology that will be applied to the GHG assessment. It defines why the specific data is relevant, how it will be measured, and what categories or frameworks are used.
General information
What is the difference between On-premise and Externalized?
We differentiate two types of datacenters:
On-premise: Servers located in a building or a room directly operated by the client. Emissions will then be split between scope 1, 2 and 3
Externalized: Servers rented through an external partner. You usually rent equipment for a month or a year, and use it as much as you want. Emissions fall under scope 3.1 (service purchases - PHYSICAL_SERVER), even for electricity or refrigerants.
Why consider emissions from outsourced and on-premise datacenters?
This emission post can easily be the most important for tech companies using this type of servers. It is also an area with a huge potential of reduction
Don’t hesitate to also mention what’s taken into account and what’s not (e.g., for food: The impact of a food purchase takes into account emissions related to its whole life cycle, from agriculture to consumption.)
Type of raw data
The analysis is split in 4 steps:
Electricity consumption of the datacenters: This consumption is estimated using the data provided by the client (kWh + country). The client also precise if the consumption they share includes the PUE or not - If not, we just need to multiply the power consumption shared by it.
Data transfer: Here we estimate the electricity consumption linked to data flowing through the DC (in and out). We collect the amount of GB transferred, the type of transfer (internet or between DCs) and use hypothesis to convert it in kWh.
Air conditioning: This emissions post is often estimated with a proxy, using the electricity consumption of the servers. We can still use the amount of refrigerant gas refilled (in kg).
Equipment inventory: This is a “classical” IT inventory, but focused on certain types of equipment - Server, Switch, Router, Firewall - so we just need to map the inventory of the client to our database.
Processing between methodologies
The emissions aren’t computed exactly the same way between GHG P and BEGES:
GHG P: The hardware impact isn’t amortized. Equipment that is bought the year of the assessment is fully counted in the analysis. If we don’t have the year of purchase, we count 1/3 of the equipment emissions
BEGES: The hardware impact is amortized over 3 years. Equipment bought more than 3 years ago aren’t counted in the analysis.
Purchase of renewables cannot be handled directly in this analysis. If you client has on-premise servers (for which electricity would go under scope 2), you’ll need to use the building module to differentiate market-based and location-based emissions
Types of emission factors
Electricity: IEA emission factors
Hardware amortization: Based on our hardware database, listing a variety of servers (mainly Dell and HP). This list has been formerly built by Boavizta, and comes from suppliers LCAs
Refrigerant gas leaks: Using data in g/MWh from the ADEMExARCEP study, and the gas r-134-a as reference.
Data transfer: Only electricity consumption of the network is taken into account. to convert the GB transfered in kWh, we’re also using the ADEMExARCEP study figures.
Others questions
Why are emissions from outsourced datacenters counted as service purchases instead of upstream leased assets?
Outsourced datacenters are sold by an external provider which also does all the maintenance, management of the whole building, and all the work around the servers you are renting. That’s why we assume it is closer to a service purchase than just a leasing.
My client’s supplier is buying 100% renewable power for their datacenter, can we take it into account?
If you client is using outsourced data centers (= emissions are accounted fully under scope 3), we do not use market-based emission factors for consistency reason (all our scope 3 emission factors are location based).
One important thing to note: a datacenter is never using directly 100% renewable energy, they are buying renewable energy contracts to match their electricity purchases. In reality, data centers are usually using an electricity more carbonated than the average of the grid due to their on-site generators.
This rule isn’t explicitly written in the GHG Protocol, and the use of market-based EFs for scope 3 is not forbidden.