Skip to main content

Get Started

1. Creating a Volume​

Follow these steps to create a new volume from the RAIC console.

  1. Navigate to Storage > Volumes in the left-hand navigation menu.
  2. Click Create Volume.
  3. Enter a descriptive name for the volume (e.g., training-data-vol-01, checkpoint-storage).
  4. Select the location where you want to provision the volume.
  5. Specify the Vvolume size. Enter a value between 1 GiB and 100 TB. The maximum available size may be limited by your remaining quota at the selected location.
  6. Review the configuration and click Create.

The volume will be provisioned and appear in your Volumes list with a Ready status once creation is complete.

tip

Use a consistent naming convention for your volumes to make it easier to identify them when attaching to clusters. For example, include the project name and data type: project-alpha-training-data or llm-finetune-checkpoints.

2. Attaching a Volume to Supercomputer Cluster​

Volumes are attached to Supercomputer clusters during the cluster creation process. At the moment you can map one volume to specific mount paths on your cluster.

  1. Navigate to Supercomputers on the main navigation and begin the cluster creation workflow.
  2. Complete the cluster configuration (node type, quantity, location, etc.) as usual.
  3. In the Storage section of the cluster creation form, choose Add Volume.
  4. Select a volume from the dropdown list. The dropdown displays all available volumes at the same location as the cluster you are creating
  5. Specify the mount path where the volume should be accessible on the cluster (e.g., /mnt/data).
  6. Complete the remaining cluster configuration and click Launch.

Know more about Supercomputers here.

Once the cluster is provisioned, the attached volumes will be mounted and accessible at the specified paths.

note

The volume dropdown only shows volumes that are provisioned at the same location as the cluster being created. If you do not see the volume you need, verify that the volume was created at the correct location.

Best Practices​

1. Volume Sizing​

Right size your volumes based on actual data requirements. Provisioning unnecessarily large volumes consumes quota that could be used elsewhere.

2. Naming Conventions​

Adopt a consistent naming scheme to keep your volumes organised as usage scales. A recommended pattern includes the project name, data purpose, and an optional sequence number.

Example names: llm-finetune-training-data (Training dataset for LLM fine-tuning project) vision-model-checkpoints-01 (Checkpoint storage for vision model, first volume)

3. Quota Monitoring​

  • Regularly review your volume usage per location to avoid unexpected quota exhaustion during critical workflow periods.
  • Submit quota increase requests proactively if you anticipate scaling up storage needs for new projects or larger datasets.
  • Delete unused volumes to free up quota for new provisioning needs.

Contents