Alauda Container Platform

GPU

Use this page when you want to run container workloads on GPUs or choose the right GPU product path in ACP.

Use the NVIDIA vendor integration for hardware enablement, direct GPU allocation, DRA, GPU Operator, or NVIDIA metrics. Use HAMi when the user task starts from sharing, virtualization, or scheduling behavior built on a GPU base.

TOC

Choose a GPU pathContainers and VMsRelated pages

Choose a GPU path

NeedTypical path
Use NVIDIA GPUs with vendor-standard componentsNVIDIA vendor integration
Share, virtualize, or schedule GPU resources across workloadsHAMi sharing and scheduling path
Use claim-based GPU allocationDRA, when the target product delivery explicitly supports a DRA driver
Use GPU passthrough for virtual machinesACP Virtualization virtual machine creation documentation

Containers and VMs

Container workload GPU usage and virtual machine GPU passthrough are different paths.

  • Container workload GPU usage is covered by this accelerator section and related product sites.
  • If a GPU resource is allocatable on the node but not visible in ACP quota pages, register the accelerator quota field metadata in Accelerator resource quota.
  • Virtual machine physical GPU passthrough is configured in ACP Virtualization documentation.