Storage Namespace

Overview

A Storage Namespace provides the storage backend used for hosting GenAI model artifacts such as model weights, tokenizer files, and related metadata. Storage namespaces allow models to be loaded efficiently into GPU memory and avoid repeated downloads from remote repositories.

Each model that uses customer-hosted or customized artifacts must be associated with a storage namespace so the system knows where the model files are stored.

Using a storage namespace eliminates the need to download large model files (which can take 30–60 minutes and be error-prone) during deployment.

---

Why Storage Namespaces Are Needed

Storage namespaces are used to:

• Host custom or fine-tuned models created by partners or enterprises
• Avoid repeated downloads from public repositories such as Hugging Face
• Improve deployment reliability and model load performance
• Provide control over where model artifacts are stored and managed

Storage namespaces are typically backed by customer-managed object storage infrastructure.

Info

Enterprise deployments may choose to use public cloud object storage such as AWS S3 in addition to on-premises storage platforms.

---

Supported Storage Backends

Storage namespaces support multiple backend types, depending on customer infrastructure, including:

• AWS S3
• Ceph
• VastData
• Weka
• Azure Storage
• Google Cloud Storage
• MinIO

Note

Currently, AWS S3 is the recommended and validated option. Support for additional backends expands based on customer-provided infrastructure.

---

Create Storage Namespace

1. Navigate to Operations Console → GenAI → Storage Namespaces
2. Click New Storage Namespace
3. Provide the following details:
4. Name — A unique name for the storage namespace
5. Description — Optional description

1. Select the Storage Provider Type configured in the environment.

1. Enter the required storage credentials for the selected backend.

Example (AWS S3): - Bucket Name
- Region
- Access Key
- Secret Key

1. Click Save Changes

After saving, the storage namespace appears in the list and becomes available during Model Creation, allowing selection of the storage location for model artifacts.

---

Access Keys

Access keys are required to authenticate with the configured storage backend when uploading model artifacts. These keys are used when configuring tools such as the AWS CLI to upload model files into the storage namespace.

Info

A maximum of two access keys can be active at any given time.

Create Access Key

• Navigate to the Storage Namespace
• Select the Access Keys tab
• Click Create New Key

A new access key is generated automatically.

Delete Access Key

• Select Delete under actions for an existing access key

If two keys already exist, one must be deleted before creating a new key.

---

View All Storage Namespaces

To view all configured storage namespaces:

• Navigate to Operations Console → GenAI → Storage Namespaces

This page displays the list of all storage namespaces.

---

Delete Storage Namespace

To delete a storage namespace:

1. Navigate to Operations Console → GenAI → Storage Namespaces
2. Select the ellipsis (⋮) under Actions for the storage namespace
3. Click Delete
4. Confirm the deletion when prompted

Deleting a storage namespace removes the configuration from the system. Model artifacts stored in the underlying storage backend are not deleted automatically.

---