Skip to content

GPU Quota Management

This section describes how to use vGPU capabilities on the DCE 5.0 platform.

Prerequisites

The corresponding GPU driver (NVIDIA GPU, NVIDIA MIG, Days, Ascend) has been deployed on the current cluster either through an Operator or manually.

Procedure

Follow these steps to manage GPU quotas in DCE 5.0:

  1. Go to Namespaces and click Quota Management to configure the GPU resources that can be used by a specific namespace.

  2. The currently supported card types for quota management in a namespace are: NVIDIA vGPU, NVIDIA MIG, Days, and Ascend.

  3. NVIDIA vGPU Quota Management: Configure the specific quota that can be used. This will create a ResourcesQuota CR.

    - Physical Card Count (nvidia.com/vgpu): Indicates the number of physical cards that the current pod needs to mount. The input value must be an integer and **less than or equal to** the number of cards on the host machine.
    - GPU Core Count (nvidia.com/gpucores): Indicates the GPU compute power occupied by each card. The value ranges from 0 to 100. If configured as 0, it is considered not to enforce isolation. If configured as 100, it is considered to exclusively occupy the entire card.
    - GPU Memory Usage (nvidia.com/gpumem): Indicates the amount of GPU memory occupied by each card. The value is in MB, with a minimum value of 1 and a maximum value equal to the entire memory of the card.
    

Comments