Goal

What is our purpose in implementing this? What are we enabling by doing this work? Time-box goals to 4-6 months.

This report provides a comprehensive technical overview of how the Kubernetes Network Driver model relates to scheduling and workload configuration.

Benefit Hypothesis:

This work will explain the limitations of Device Plugins and the challenges this feature poses for complex workloads. DRA (Device Resource Allocation) addresses these issues by enabling more dynamic, topology-aware scheduling, supporting advanced scenarios like GPU/NIC alignment and multi-node AI workloads.

This effort will benefit customers and OpenShift Networking in several ways:

• Adopt a Kubernetes-native networking model
• Enable a high-performance networking enablement
• Support backward compatibility and coexistence with Multus CNI
• Improved consistency and extensibility
• Future-proof OpenShift networking

Resources

A document will be created that will expose interactions with CNI and Device Plugins and identifies potential improvements through a combined DRA + NRI approach.

The report also reviews the current state of CNI (Container Network Interface) usage and emerging APIs such as NRI (Network Resource Interface) for workload configuration. These APIs provide mechanisms for defining and enforcing network resources in a Kubernetes environment, but integration with scheduling is still evolving.

Finally, the document highlights the SIG-Network subproject on Multi-Network, which improves the user-facing API—such as through constructs like PodNetwork and describes how it interacts with DRA to deliver a consistent, flexible, and multi-network-aware experience for Kubernetes workloads.

Responsibilities

The bulk of the effort will be from OpenShift's Core Networking team, and supported by their upstream interactions in the community. 

Success Criteria

The document should present a clear explanation of foundational and related technologies, and provide a clear recommendation of OpenShift Networking's path forward in this space to inform future-created Jira Features. 

Results

Add results here once the Initiative is started. Recommend discussions & updates once per quarter in bullets.