Autoscaling containers

[emerson-argueta]

There was a discussion earlier about dynamic load: #78

I have a question similar to this but different in handling dynamic load.

To premise, with Kamal, you can specify many servers to handle the load for a service. Then, in your infrastructure, you can set up a load balancer to handle requests to the service. Having this is great because you can scale by adding more VMS. The way Kamal works is by starting up one container per server for your service. (correct me if I'm wrong)

Now, my question: Would it be helpful to have more than one container for a service per server? For example, for a service tied to 4 of my servers, there could be 1 to any containers spun up in this group of servers depending on the CPU consumption across the group. By doing so, it could save resources on my servers.

I'm unsure whether a feature like this is needed for Kamal, but I would like to hear anyone's thoughts on this, critiques on what I need help understanding, etc.

[emerson-argueta]

This was partially answered in #16
However, this is more of a manual scaling. Perhaps, the amount of workers on a puma server could be adjusted automatically depending on load?