Creating a GPU cluster
This section explains how to create GPU clusters.
By default, the cloud has a zero quota for creating GPU clusters. To change the quota, contact technical support
After creating a cluster, you can add VMs to it.
-
In the management console
, select the folder for a new GPU cluster. -
Select Compute Cloud.
-
Select GPU clusters.
-
Click Create a cluster.
-
Specify the cluster name:
- The length can be from 3 to 63 characters.
- It may contain lowercase Latin letters, numbers, and hyphens.
- The first character must be a letter. The last character can't be a hyphen.
-
If needed, add a description to distinguish the clusters.
-
Select an InfiniBand fabric to create the GPU cluster in:
-
fabric-1
: Use for creating GPU clusters with NVIDIA® H100 NVLink with Intel Sapphire Rapids (Type A) and NVIDIA® H100 NVLink with Intel Sapphire Rapids (Type B) VMs. Both types could be mixed in the cluster. -
fabric-4
: Use for creating GPU clusters with NVIDIA® H100 NVLink with Intel Sapphire Rapids (Type C) VMs.
For more details, see InfiniBand fabrics.
-
-
Click Save.
You can also create a GPU cluster while creating the first VM in it. Under Computing resources, in a GPU cluster, click Create button and fill in the fields.
If you don't have the Nebius AI command line interface yet, install and initialize it.
The folder specified in the CLI profile is used by default. You can specify a different folder using the --folder-name
or --folder-id
parameter.
-
View the description of the command that creates a GPU cluster:
ncp compute gpu-cluster create --help
Please note that you can currently create clusters with the InfiniBand connection type only.
-
Create a cluster:
ncp compute gpu-cluster create --interconnect-type infiniband
The cluster will be created in the default InfiniBand fabric. To make sure the fabric suits your VMs' platform, add the
--interconnect-physical-cluster
parameter. For example, if you are going to add NVIDIA® H100 NVLink with Intel Sapphire Rapids (Type C) VMs to the cluster, selectfabric-4
:ncp compute gpu-cluster create --interconnect-type infiniband \ --interconnect-physical-cluster fabric-4