Alauda Container Platform

Install

TOC

Installing Kubernetes Hardware accelerator Toolkit

Prerequisites

Before installation, ensure the following requirements are met:

  1. Kubernetes Cluster: Version ≥1.25 with DevicePlugins feature gate enabled.
  2. NVIDIA Drivers: Installed on all GPU nodes . Verify with nvidia-smi.
  3. Container Runtime: Configured with NVIDIA Container Toolkit (≥1.7.0) for GPU support.

Installing via Web Console

  1. Navigate to Cluster Plugins:

    • Go to Platform Management → Catalog → Cluster Plugin
    • Search for "gpu" and click Install

  2. Feature Toggles: Enable/disable advanced capabilities during installation:

    OptionFunctionalityRecommended Scenario
    PGPUPhysical GPU isolation with NUMA-aware schedulingAI training/high-performance computing
    vGPUVirtual GPU slicing via GPU-ManagerMulti-tenant sharing/resource quotas
    MPSMulti-Process Service for compute/memory sharingLow-latency inference/parallel tasks