Epic: Support fixed bigger compute units

[stradig]

Motivation

We got a couple of high priority customer requests to support bigger compute units than we currently support (8 CUs). Supporting that would allow onboarding a class of high value customers which needs those big computes.

Supporting this in the general case including autoscaling is a larger task. We expect to run into various issues when scaling from a very small to a very big instance. Some of those issues should be solved by work planned for later this year. This includes but is not limited to scheduling issues (would be easier with automatic live migration), difficulties upscaling to the requested size (would be easier with a planned new scaling strategy) and more.

It would be relatively simple to support fixed size large instances. Those would be provided to special customers on demand (e.g. by enabling a feature flag for them). They would probably have a number of caveats:

• No autoscaling
• Startup and/or reaching the requested size could take relatively long

We believe this may provide enough value for those customers so that they would choose Neon.

DoD

For special customers we could enable fixed large compute instances on demand. The customers could then select such an instance using the UI.

[stradig]

We also need to take into account the autoscaling graphs!