GKE, Cloud SQL, Cloud Filestore and Cloud Memorystore are located in the same private VPC to make them more secure and to make GKE connect to the rest of them.
- Google Kubernetes Engine is used to run polyason. We connect to containers of polyaxon API through SSH tunnel. The permission of GKE developer is required to make a SSH tunnel.
- Google Cloud SQL is used to store resulting metrics of polyaxon and so on.
- Google Filestore is used to store uploaded files in order to build docker images.
- Google Cloud Storage
- data: A bucket is used to store data uploaded by polyaxon jobs.
- output: A bucket is used to store outputs uploaded by polyaxon experiments.
- logs: A bucket is used to store polyaxon logs.
- Google Cloud TPU can be used, when we specify TPU resources in polyaxon configurations.
- (Optional) Cloud Memorystore is used to stabilize polyaxon's queue with celery.
