nebari-dev
[ENH] - Consider making profile resource sizes and node sizes match
Feature description
With default config, the small profile uses 2 cpu and 8 GiB of memory, but the smallest node we use is 4 cpu and 16 GiB of memory. While this could make sense for large Nebari deployments, most Nebari deployments I see have relatively few users running simultaneously (like 3 or less and I suspect most of the time there is only a single user at a time).
It seems like it makes sense then to match up profile sizes with node resources to use costs most efficiently on the user nodes.
Value and/or benefit
Cost Savings
Anything else?
No response
That makes sense; the original idea of keeping a somewhat larger instance type than the group itself was that we could balance more user pods in the same node without always requiring to scale up. For example, if a second user logs in and spawns an instance, it would be (depending on the chosen profile) assigned to the same node, which has some benefits:
- All the images are already downloaded and present in the node
- No need for scaling times
While this could make sense for large Nebari deployments, most Nebari deployments I see have relatively few users running simultaneously (like 3 or less and I suspect most of the time there is only a single user at a time).
It is the case most of the time; I think this can be discussed in our next community meeting, as it is mostly a tradeoff/balance thing.
