[ENH] - Lower cost node sizes on GCP Nebari

[Adam-D-Lewis]

Feature description

Note: Prices are at the time of writing and for the us-central-1 region.

The general node on GCP has 8 cpus and 30G of memory (n1-standard-8). This node size costs ($0.38/hr = $275/month), but the most of the time the cpus are doing very little and GCP has the ability to create custom node sizes at the cost of $25/cpu-month and $3/GB-month of RAM (up to 6.5GB/cpu) + $7/GB-month for additional RAM (beyond 6.5GB/cpu).

E.g. If we could get away with a 4 cpu 30GB of RAM node then the cost would be $205/month.

Value and/or benefit

25% Lower fixed costs for using Nebari. It's not clear to me what the ideal resources are for the node group, but it seems likely that we can reduce the cpu amount from 8 to 4 without ill effects. According to Grafana, in the last 7 days, this particular cluster never went beyond 1 cpu used for the general node, and the vast majority of the time needed less than 1 cpu.

Alternatively a e2-highmem-4 "cost optimized" instance with 4 cpus and 32GB costs ~$130/month.

Anything else?

More info on how to create custom machine types at https://cloud.google.com/compute/docs/instances/creating-instance-with-custom-machine-type

[Adam-D-Lewis]

Looking further, it seems easy to change the default node type to E2 instances on GCP and reduce costs by ~50%