The following table shows our recommended quota values. We recommend
setting the quota values as a function of the number of concurrent
training jobs ( num_concurrent_pipeline
) and the number of CPUs in the
requested region. The recommended values are valid only if you use the
default Compute Engine resource configuration for your workflow.
| Service | Quota | Recommendation |
|---|---|---|
|
Compute Engine API
|
CPUs | num_concurrent_pipeline
x 440 CPUs |
|
Compute Engine API
|
Persistent Disk Standard (GB) | num_concurrent_pipeline
x 5TB persistent disk |
|
Vertex AI API
|
Restricted image training CPUs for N1/E2 machine types per region | num_concurrent_pipeline
x 440 CPUs |
|
Vertex AI API
|
Restricted image training total persistent disk SSD storage (GB) per region | num_concurrent_pipeline
x 8TB persistent disk |
|
Vertex AI API
|
Resource management (CRUD) requests per minute per region | num_concurrent_pipeline
x 150 |
|
Vertex AI API
|
Job or LRO submission requests per minute per region | num_concurrent_pipeline
x 6 |
