Namespaces

Definition

Namespaces in a ML Serving project are organization entities for your models. It allows user to simply classify their models depending on their needs by attaching them to an existing namespace.

Each namespace is linked to an object storage container of the same public cloud project where users can upload their serialized machine learning models. That object storage container is configurable.

Each namespace is also linked to a docker registry. The default docker registry is an OVHcloud provided one but this can be configured.

Considerations

• Number of created namespaces is restricted to 100 maximum per public cloud user.
• Each namespace is linked to one and only object storage container of the same public cloud project.
• Each namespace is linked to one and only docker registry.
• Each namespace can contain as many models as the user wants.
• Each model deployed inside a namespace will be reachable from the same base url such as https://<id-of-namespace>.<cx>.gra.serving.ai.ovh.net where <id-of-namespace> will be a generated unique identifier of the namespace and <cx> the identifier of your assigned cluster.

Under the hood

Namespaces in a ML Serving project correspond to namespaces in a kubernetes cluster.

Going further

• You can check the OVHcloud documention on how to create a namespace.

These guides might also interest you...