Connect your existing Container Service for Kubernetes (ACK) cluster to Platform for AI (PAI) to use as a compute resource for Data Science Workshop (DSW) development and Deep Learning Containers (DLC) training jobs.
Prerequisites
Contact your PDSA to complete the following preparations:
An ACK cluster is ready for use in PAI.
Your account is on the allowlist.
Configure self-managed compute resources
Step 1: Create a resource group
First, create a resource group in PAI and map it to the specified ACK cluster.
Log in to the PAI console. In the upper-left corner, select the region where your ACK cluster is located.
In the navigation pane on the left, choose AI Computing Resources > Resource Pool.
Click the Self-managed compute tab, and click Create Resource Group. In the dialog box, select your ACK cluster from the drop-down list.
Step 2: Add a resource quota
After you create a resource group, configure a resource quota for it. This quota provides resource isolation and management without incurring fees.
For detailed instructions, see Add a resource quota.
On the Resource Quota page, click the Self-managed compute tab, and click Add a resource quota in the upper-left corner.
Use self-managed compute resources
After completing the configuration, you can use the self-managed compute resources in different PAI components.
Use in DSW instances
In the Resource Information section, set Resource Type to Resource Quota, and then select the resource quota for your self-managed compute resources from the drop-down list.
The Resource Overview section then displays the GPU type (for example, V100) and an Intelligent Policy tag. The specifications table below lists the quota details, such as specification name, total number of nodes, GPU model, number of GPUs, vCPU cores, and memory.
Use in DLC
When you create a DLC job, set Resource Type to Self-managed compute and Source to Resource Quota. Then, select the resource quota for your self-managed compute resources from the drop-down list.
The interface then displays the specifications of the selected resource quota, including the specification name, GPU model, number of GPUs, vCPU cores, and memory.
Resource usage
You can view the usage of your self-managed compute resources on the Resource Overview page, located on the Overview tab.
Click the Self-managed compute tab to view details about your self-managed compute resources, such as name, status, number of nodes, and number of GPUs.