[discuss]: creating cloud organizations for communities and cloud resource limits

[choldgraf]

Context

When deploying a recent hub for the first time, we ran into some issues with resource limit quotas on AWS when requesting more GPUs:

• https://github.com/2i2c-org/infrastructure/issues/1309

The issue arose because we had a relatively new AWS organization created for the community/event, I think with the goal of simplifying our billing process a bit. However, it's clear that creating new organizations also comes with the risk that those organizations have caps on the resources available to them. This might be a problem for certain communities.

Questions

• When / why do we want to create a new AWS organization for communities instead of re-using an existing one?
• When will the resource limit quotas be a blocker on doing this?
• Are there any ways around the resource limit issue that still lets us create new AWS organizations easily?
• Do similar challenges / solutions exist for other cloud providers?

[jmunroe]

Some questions:

• Does our infrastructure require an independent AWS organization for each Hub deployment?
• Can multiple AWS organizations share the same AWS billing account?
• Are resource limits associated with AWS organizations or AWS billing accounts?

Especially for hubs that are event-based, I think it makes sense to have those hubs live an 2i2c-owned organization (whether that be AWS, GCP, Azure, etc.)