Assigned
Status Update
No update yet.
Comments
gu...@google.com
The AI Platform engineering team is aware of this request and are investigating implementations. There is no ETA at this time for a release, but all further updates should occur here.
Description
What you would like to accomplish:
Allow new n1 machines to allow a default of 0 nodes such that the auto-scaling algorithm can shut down all nodes when no predictions are being made. This would reduce monthly costs for online predictions
How this might work:
Similar to how the current legacy (mls1) machines allow for a default of 0 nodes.
If applicable, reasons why alternative solutions are not sufficient:
Alternative solution would be to delete the model version after predictions are done. This, however, is a tedious workaround and requires additional unnecessary steps.