InfiniBand Link Aggregation via Virtual Address Mapping

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Link aggregation (LAG) protocols like LACP are not directly applicable in InfiniBand (IB) networks, which lack a centralized management system like the subnet manager (SM) and subnet management agents (SMAs) present in ethernet networks, necessitating a different approach for establishing, modifying, and monitoring IB LAG.

Innovation Solution

Utilizing an SM and SMAs to manage IB LAG by mapping virtual network addresses to physical IB ports, enabling load-balancing and traffic management through configuration of internal forwarding tables in IB switches, and employing MAD messaging for communication between SM, SMAs, and IB daemons.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If LACP protocol is used for link aggregation management, then load-balancing and port monitoring are enabled, but the protocol cannot be directly applied in InfiniBand networks due to lack of centralized management system

Engineering Contradiction:
Improvelink aggregation managementVSAvoidprotocol compatibility
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent introduces MAD (Management Datagram) messaging as an intermediary mechanism to enable LACP-like functionality in InfiniBand networks. The SM (Subnet Manager) and SMA (Subnet Management Agent) act as mediators that translate and adapt LACP operations to the InfiniBand management architecture, allowing link aggregation to function across different network protocols without requiring LACP native support in InfiniBand

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent makes the InfiniBand SM and SMA multi-functional by enabling them to perform both traditional InfiniBand management tasks and LACP link aggregation management. This universal approach allows a single centralized management system to handle multiple protocols and functions, eliminating the need for separate LACP implementation in InfiniBand networks

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

2Productivity

If virtual network address mapping is implemented, then load-balancing and traffic management are enabled, but forwarding table configuration complexity increases

Engineering Contradiction:
Improveload-balancing efficiencyVSAvoidforwarding table configuration
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements self-service mechanisms where the SM automatically generates and distributes virtual network address mappings to switches, and the switches autonomously configure their forwarding tables based on received MAD messages. This automation eliminates manual forwarding table configuration, reducing complexity while maintaining load-balancing capabilities through dynamic virtual address mapping

Inventive Principle:
Principle #25Self-service

3Reliability

If centralized SM and SMA management is used, then IB LAG can be established and monitored, but communication overhead between components increases

Engineering Contradiction:
ImproveLAG management capabilityVSAvoidcommunication overhead
Core Design Contradiction:
ReliabilityVSLoss of energy

Solution Approach 1:

The patent implements periodic MAD messaging between SM, SMAs, and IB daemons to monitor and maintain LAG state. Instead of continuous communication, the system uses periodic status checks and event-driven updates, reducing communication overhead while ensuring reliable LAG management and monitoring through structured periodic interactions

Inventive Principle:
Principle #19Periodic action

Data Source

PatentUS20250379813A1Link aggregation in infiniband networks
Publication Date: 2025.12.11 MELLANOX TECHNOLOGIES LTD(IL)
  • US20250379813A1 patent drawing
  • US20250379813A1 patent drawing
  • US20250379813A1 patent drawing

AI summary

Systems and methods herein are for one or more processing units of a subnet manager (SM) to communicate configuration information with at least one subnet management agent (SMA) that is associated with at least one switch and with a host machine, the configuration information to enable the at least one switch to configure a forwarding table based in part on a mapping of at least one virtual network address to physical network addresses of two or more physical ports of the host machine, and the configuration information to enable the host machine to communicate with other host machines using the at least one switch and the at least one virtual network address.