Cross-Card VF Link Aggregation for DPU Failover

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing DPU systems face a single point of failure issue due to all virtual function ports (VFs) being connected to a single DPU card, leading to network interruptions upon hardware failure, compromising high availability.

Innovation Solution

Implement cross-card link aggregation of virtual function ports by establishing communication links between multiple DPU cards and virtual machines, enabling message transmission through multiple paths to avoid single point failures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If only one DPU card is used for virtual machine communication, then device complexity is reduced, but reliability deteriorates due to single point of failure

Engineering Contradiction:
Improvenetwork availabilityVSAvoidDPU card configuration
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent combines multiple DPU cards into a unified communication system through link aggregation. Multiple physical links are merged into a single logical connection between virtual machines and switches, allowing traffic to be distributed across multiple DPU cards. This merging approach improves reliability by eliminating single point of failure while managing complexity through unified configuration and control of the aggregated links.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent segments network traffic into different virtual function ports (VFs) that can be distributed across multiple DPU cards. Each VF is independently configured and can be assigned to different physical DPU cards, enabling fault isolation and load distribution. This segmentation allows the system to maintain reliability through redundancy while managing complexity through modular VF configuration.

Inventive Principle:
Principle #1Segmentation

2Reliability

If multiple DPU cards are aggregated for high availability, then reliability is improved, but device complexity increases

Engineering Contradiction:
Improvefault toleranceVSAvoidlink aggregation configuration
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements universal functionality across multiple DPU cards through standardized link aggregation protocols. The same configuration and control mechanisms apply to all DPU cards in the aggregation, allowing them to function interchangeably. This universality simplifies management despite having multiple cards, as the system treats all members of the aggregation uniformly through standard protocols and unified control planes.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent introduces intermediary components such as virtual switch controllers and link aggregation managers that mediate between multiple DPU cards and the network. These intermediaries handle the complexity of coordinating multiple cards, managing link negotiation, and distributing traffic, thereby shielding the user from the underlying complexity while maintaining high availability through redundant DPU configurations.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Reliability

If traffic passes through a single DPU card, then device complexity is minimized, but loss of information occurs during hardware failure

Engineering Contradiction:
Improvenetwork continuityVSAvoidcommunication path configuration
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary actions by pre-configuring multiple DPU cards and establishing redundant communication paths before any failure occurs. Link aggregation is set up in advance with multiple active paths, and the system maintains standby configurations that can immediately take over upon failure detection. This preliminary preparation ensures network continuity without requiring complex real-time reconfiguration during failure events.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies local quality by enabling independent configuration and failure isolation at the virtual function port level. Each VF can be independently assigned to specific DPU cards, allowing localized failure containment where only affected VFs are isolated while other VFs continue operating through alternative DPU cards. This local quality approach maintains network continuity for unaffected services while managing the complexity of multi-card configuration through granular, service-specific routing decisions.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4407451B1Method for cross-card link aggregation of virtual function ports of data processing units, device and medium
Publication Date: 2026.05.13 YUSUR TECH CO LTD
  • EP4407451B1 patent drawingFigure 1
  • EP4407451B1 patent drawingFigure 2
  • EP4407451B1 patent drawingFigure 3

AI summary

The present disclosure relates to a method and a device for cross-card link aggregation of virtual function ports of DPUs, and a medium. The method includes: providing a plurality of data processing units, each of which is communicatively connected to a virtual machine and a switch respectively, performing link negotiation for the virtual machine and the switch respectively, acquiring a to-be-communicated data message, further determining a communication link for the to-be-communicated data message, where the communication link is a message channel between a virtual function port in the virtual machine, a virtual function port representor corresponding to the virtual function port in the virtual machine, an uplink corresponding to a physical port in the data processing unit, the physical port in the data processing unit, and the switch; and performing communication for the to-be-communicated data message based on the communication link. Because the plurality of data processing units is provided, and each of the data processing units is communicatively connected to the virtual machine and the switch respectively, the message can be communicated through a plurality of communication links, thereby avoiding a problem of network interruption caused by single point of failure, and ensuring high availability.