Guidelines Feedback - Impact of reservation rate on embodied carbon calculation

Hi!

I was wondering should the reservation rate of servers be included in the calculation of application share of embodied carbon in the guide?

The [current formula](https://sci-guide.greensoftware.foundation/M/), `M = TE TS RS`, implicitly assumes that a server is always reserved for use. For instance, if a server is reserved for use on average only 50% of its lifetime, then the sum of all usage calculations using this formula would only account for half of the server’s true embodied carbon.

A more accurate formula would be `M = TE TS RS / ARR`, where `ARR` represents the average reservation rate (0–1) of that type of instance in the data center over the lifespan of the server. While cloud providers do not publish reservation rates, estimates could be made. These rates may vary based on instance type (e.g., general-purpose vs GPU vs large memory instances), data center location and provider.

Large cloud providers like AWS offer spot / preemptible instances at high discounts to drive up the reservation rate while still maintaining capacity for on-demand instances. If the utilization rate is sufficiently close to 100% then this parameter can be reasonably ignored. I would assume smaller providers and private cloud setups can have significantly lower reservation rates.