You can compare total better than per user at these scales.
Lemmy needs a certain amount of performance to keep up with federation, but once you have all the images and posts and comments you don’t need second versions until you scale to a size that mandates multiple machines. Which I would guess is more in the 6+ digit user range, where you start averaging requests per second not minute.
In some sense, every lemmy user is a user of your instance via federation. You need to pay the performance for all 100k of us whether your instance has 10 or 10k of those. Local users are just a bit extra demanding on your hosting resources.
I suspect the bias we see here with larger instances paying a bit more (50-ish instead of 10-ish) is more due to reliability and snappyness than actual performance needs too. You tend to get optional smaller-gains pricier perks you might not go for for a smaller instance.
You can compare total better than per user at these scales.
Lemmy needs a certain amount of performance to keep up with federation, but once you have all the images and posts and comments you don’t need second versions until you scale to a size that mandates multiple machines. Which I would guess is more in the 6+ digit user range, where you start averaging requests per second not minute.
In some sense, every lemmy user is a user of your instance via federation. You need to pay the performance for all 100k of us whether your instance has 10 or 10k of those. Local users are just a bit extra demanding on your hosting resources.
I suspect the bias we see here with larger instances paying a bit more (50-ish instead of 10-ish) is more due to reliability and snappyness than actual performance needs too. You tend to get optional smaller-gains pricier perks you might not go for for a smaller instance.