CML utilizes the CDP Control Plane to manage Data Services so you can provision and delete CML Workspaces. CDP Control Plane leverages cloud native capabilities to dynamically access CPU, memory, and GPU resources along with cloud-managed Kubernetes (K8s) to provide the infrastructure.

During provisioning the Machine Learning application is configured to authenticate end users of the service (ML) via the CDP identity provider, which is “chained” back to the customer’s identity provider. As a result, Machine Learning provisioned instances allow for seamless customer SSO.

When you provision a CML workspace, the following happens:

  • CDP Control Plane performs the following in the cloud environment:
    • Requests a TLS Certificate and domain name with the domain
    • Identifies the SSO configuration
    • Identifies the SDX configuration for the environment
  • CDP Control Plane provisions a managed Kubernetes cluster
  • CDP Control Plane installs CML into the Kubernetes environment
  • Storage is mounted directly via managed service providers

CML uses the cloud provider load balancer and networking infrastructure to partition the resources. CML also leverages the cloud provider infrastructure to enable the customer to specify autoscaling.

CML provisions the DNS and the certificate. CML renews the certificates for the customer on an ongoing basis.