Features of Kubeflow on GCP

Reasons to use Kubeflow on Google Cloud Platform (GCP)

Running Kubeflow on GCP brings you the following features:

  • You useDeployment Manager todeclaratively manage all non-Kubernetes resources (including the GKEcluster). Deployment Manager is easy to customize for your particular usecase.
  • You can take advantage ofGKE autoscaling to scaleyour cluster horizontallyand vertically to meet the demands of machine learning (ML) workloads withlarge resource requirements.
  • Cloud Identity-Aware Proxy (Cloud IAP)makes it easy to securely connect to Jupyter and otherweb apps running as part of Kubeflow.
  • Kubeflow’s basic authentication service supports simple username/passwordaccess to your Kubeflow resources. Basic auth is an alternative to CloudIAP:
    • We recommend Cloud IAP for production and enterprise workloads.
    • Consider basic auth only when you want to test Kubeflow and use itwithout sensitive data.
  • Stackdriver providespersistent logs to aid in debugging and troubleshooting.
  • You can use GPUs and Cloud TPU toaccelerate your workload.

Next steps