So I had a deeper look into this and I am not sure...
# general
g
So I had a deeper look into this and I am not sure spinning an instance for each organization would work. When we have 1000 organizations, even if some of them are not used iit would require to have always on servers to handle thiis. I know you are using this strategy for Ory Cloud and I am wondering if you have some figues around that. An other solution could be to add a key iin user models for all ory component but not sure how to handle this. if we were to do a Pull Request, where should we start ?
So we made a test and ory kratos use around 7MB. t means that for 5000 users it would require 75 GB even in idle mode. Do you confirm this ?
m
Why does it mean that?
f
7*5000=35.000, not 75.000
m
Ah, sure. Why do you think it will be linear like that?
f
I also figure that if they're not used, some of that can be put into swap right? I'm not familiar enough with swap or golang to say that that's definitely the case.
@microscopic-forest-58980 I would have assumed the same thing to be honest. Is this not the case?
Just seems natural. If one car uses 1 liter of fuel per km, and you drive 1000 of them at the same time, I'd assume 1000 liter/km would be used.
m
I don't know. Have you tried with 2 users to test your linearity theory?
g
Sorry gues, did not see your answer.
Yeah actually it was 35 GB i made a typo 😕
So I assume the consumption of ressources was linear because I saw somewhere that the consumption of ressources was around 7MB of RAM so i just took a linear model to see with 5000 tenants what was the usage. I did not try to see the linear theory but will maybe see it if we want to check it more deeply. I just think that for the moment the multi-tenant with ory is maybe not suitable for us at the moment in terms of complexity (having to maintain a lot of schema on a database for iam and so on is quite resource hungry).