Overview
Learn how to add and initialize nodes, enable GPU as a service, create and manage GPU pools.
Adding GPUaaS Nodes
Adding a GPU as a Service Node
The node must have Ubuntu 24.04 pre-installed for GPUaaS
Open the add GPU as a service node wizard.
Specify a name for the node so you can easily recognise it in future
Select its region, you can create regions from the Regions menu
Enter the nodes IP address
Set a SSH port (defaults to 22).
Enter the username for accessing the node (ensure that user is able to sudo without a password).
Specify the node's role if you want it to perform any role other than Worker (which is enabled on all nodes by default)
More information about node roles is available here
SSH Key Management
After adding a node, the SSH key is available in the actions menu for the node.
Copy the key and add the copied SSH key to the authorized_keys file for the selected user on the GPUaaS node.
Click to test the connection from the admin panel to the node to validate communication.
Node Initialization
Initialize the node from the admin panel. This sets up necessary dependencies and packages using Ansible playbooks.
Enabling GPU as a Service
After initializing the controller node, enable GPU as a service. This sets up the internal Kubernetes cluster and prepares for GPU pool management.
GPU Pool Creation and Management
Adding a New Pool
Click to add a new pool specifying the following:
The region
A name for the pool (this is how end users will see it)
The GPU card(s) to assign to the pool
The time quantum (cycle time allocation for each task)
The sharing ratio (oversubscription factor)
Ansible Playbooks and Recipes
Playbook Locations and Usage
Internal Playbooks are used for node initialization, enabling GPU as a service, joining a cluster, creating pools, and managing team subscriptions. These playbooks automate the setup and configuration of your GPUaaS infrastructure and should not be altered.