Alauda Container Platform

Hardware accelerators

Use this section to choose the right accelerator path for container workloads, understand the ACP concepts that affect resource requests, and continue to the product documentation for installation, upgrade, compatibility, and troubleshooting.

Start from the scenario closest to your task. Vendor integrations prepare hardware, drivers, runtimes, and direct allocation. Sharing and scheduling products such as HAMi build on those integrations to provide sharing, virtualization, scheduling, or resource orchestration behavior.

TOC

Choose a scenarioBrowse by categoryRelated ACP pages

Choose a scenario

If you want to...Start here
Use NVIDIA GPUs with vendor-standard componentsGPU accelerator family and NVIDIA vendor integration
Run container workloads on Ascend NPUs with vendor-standard componentsNPU accelerator family and Ascend vendor integration
Share, virtualize, or schedule GPU or NPU devices across workloadsHAMi sharing and scheduling path
Use Kubernetes DRA-based resource claimsDRA mechanism
Show accelerator resources in ACP quota pages or configure quota field metadataAccelerator Resource Quota
Use GPU passthrough for virtual machinesUse Virtual Machine physical GPU configuration. This is a different path from container workload pGPU or vGPU usage.

Browse by category

  • Overview: concepts such as Device Plugin, DRA, RuntimeClass, CDI, extended resources, quotas, and monitoring entry points.
  • Accelerator families: User-facing hardware families such as GPU and NPU.
  • Products: vendor integrations such as NVIDIA and Ascend, and sharing or scheduling products such as HAMi.

For quota, plugin installation, monitoring dashboards, virtual machine GPU passthrough, and billing, continue with the corresponding ACP feature documentation: