Skip to main content

Get Started

Step 1. Creating a Supercomputer​

Follow these steps to provision a new Supercomputer from the RAIC's console.

  1. Navigate to the Supercomputers in the left-hand navigation menu.
  2. Click Create Cluster.
  3. Choose the GPU Type (H100 SXM or H200 SXM).
  4. Select the Location where you want the Supercomputer provisioned.
  5. Specify the number of Nodes you require.
  6. (Optional) Enable GPU-Direct Networking (InfiniBand) if your workload requires high-bandwidth inter-node communication for distributed training.
  7. (Optional) Attach a Volume by selecting from the dropdown of available volumes at the chosen location, and specify a Mount Path (e.g., /mnt/data).
  8. Select an OS Image from the available options.
  9. (Optional) Add a Custom Init Script to automate environment setup on each node during provisioning.
  10. Provide an SSH Key. Upload or select an existing public key that will be used to access all nodes in the cluster.
  11. Review the configuration and click Create.

Your Supercomputer will begin provisioning. Once all nodes reach Available status, the cluster is ready for use.

Step 2. Performing Node Level Actions​

Individual Node Action​

  1. Navigate to your Supercomputer's detail view.
  2. View the Nodes section to see the list of all nodes with their IP addresses.
  3. Click the actions menu (⋮) next to the target node.
  4. Select the desired action: Soft Reboot, Hard Reboot, or Reinstall.
  5. Confirm the action when prompted.

Batch Node Actions​

For operations across multiple nodes simultaneously:

  1. In the Nodes tab, select the checkboxes next to the nodes you want to act on.
  2. Click the Batch Actions button that appears above the node list.
  3. Choose the desired action from the dropdown.
  4. Review the list of affected nodes and confirm.
note

Batch actions apply the same operation to all selected nodes. If you need different actions on different nodes, perform them individually.

Step 3. Scaling Nodes​

You can scale your Supercomputer up or down by adding or removing nodes after initial creation.

  • Scale Up: Add additional nodes to increase compute capacity. New nodes are provisioned with the same configuration (GPU type, OS image, SSH key) as the existing cluster.

  • Scale Down: Remove nodes to reduce capacity and cost. Select the specific nodes you wish to remove.

Step 4. Deleting a Supercomputer​

Deleting a Supercomputer permanently removes the cluster and releases all associated resources.

  1. Navigate to the Supercomputer detail view.
  2. Click Delete
  3. Confirm the deletion.
note

Deleting a Supercomputer is irreversible. All local data on the nodes will be permanently lost. Any attached volumes will be automatically detached but will remain intact.

Contents