Host Trusted Deployment

The Host Trusted deployment scenario represents DPF for Host Trusted environments, where the DPU is treated as a host accelerator card and is managed by the host. In this deployment model, host machines function as worker nodes within the DPF management cluster, with the DPU serving as an integrated acceleration resource that enhances the host's capabilities.

This architecture allows users to leverage standard Kubernetes APIs for managing both host workloads and DPU services, creating an end to end Kubernetes-native experience.

The DOCA Management Service(DMS) is used to install and configure the DPU.

The Host trusted model supports various use cases, each optimized for specific networking and acceleration requirements.

Each use case may require tailored configurations on the DPU for optimal performance. Below are the validated user guides with corresponding deployment guides.

HBN and OVN Kubernetes

• Host Based Networking and OVN Kubernetes

OVN Kubernetes

• OVN Kubernetes on its own

HBN

• Host Based Networking on its own

DPU Passthrough

• DPU Passthrough

No DPUServices are necessary for this use case.

Before deploying DPF in Host Trusted mode, ensure you have the necessary hardware and software prerequisites in place. This includes compatible DPU hardware, host operating system configurations, and network settings. For detailed requirements, refer to the prerequisites guide.

For assistance with Host Trusted deployments:

• Review the specific use case documentation

• Check the DPF troubleshooting guide

• Consult the release notes for known issues

• Engage with the NVIDIA support team for further assistance

Once you've familiarized yourself with the Host Trusted concept, proceed to:

1. Review the prerequisites

2. Select an appropriate use case

3. Begin your DPF Host Trusted deployment