Multicast Send Duplication for InfiniBand VM Migration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

High performance computing environments face challenges with traditional network and storage bottlenecks, particularly in cloud computing architectures, where live migration of virtual machines is hindered by complex addressing and routing schemes in InfiniBand networks, leading to connectivity issues and performance degradation.

Innovation Solution

The implementation of a system and method for multicast send duplication instead of replication in a high performance computing environment, utilizing a vSwitch architecture with prepopulated or dynamic LID assignment, and dual-port virtual router configurations to provide redundant and independent network connectivity, ensuring efficient resource utilization and fault tolerance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional replication is used for multicast send in InfiniBand networks, then connectivity and data delivery are ensured, but network performance degrades and bottlenecks occur due to complex addressing and routing schemes

Engineering Contradiction:
ImproveconnectivityVSAvoidnetwork performance
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent uses multicast packet duplication where a single multicast packet is copied and sent through multiple network paths simultaneously. This approach maintains reliable data delivery to multiple receivers while avoiding the performance degradation associated with traditional replication methods that require complex addressing and routing schemes. The duplication occurs at the source node, simplifying the network infrastructure.

Inventive Principle:
Principle #26Copying

2Adaptability or versatility

If virtual machine migration is performed in InfiniBand networks, then resource allocation flexibility is improved, but downtime increases due to complex addressing and routing schemes

Engineering Contradiction:
Improveresource allocationVSAvoiddowntime
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent implements prepopulated Local Identifier (LID) assignment where multicast group memberships and routing information are established in advance before virtual machine migration occurs. This preliminary configuration allows the network to quickly adapt to migration events without requiring complex real-time addressing and routing calculations, thereby reducing migration downtime while maintaining resource allocation flexibility.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If redundant network connectivity is implemented, then fault tolerance is improved, but network complexity increases

Engineering Contradiction:
Improvefault toleranceVSAvoidnetwork complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent combines multiple network paths into a unified multicast transmission approach where a single multicast packet is simultaneously transmitted through redundant network routes. This merging of paths at the source node provides fault tolerance through redundant connectivity while avoiding the complexity that would arise from managing separate redundant transmission systems. The network infrastructure is simplified compared to traditional replication methods.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS11936556B2System and method for a multicast send duplication instead of replication in a high performance computing environment
Publication Date: 2024.03.19 ORACLE INT CORP
  • US11936556B2 patent drawing
  • US11936556B2 patent drawing
  • US11936556B2 patent drawing

AI summary

Systems and methods for multicast send duplication instead of replication in a high performance computing environment. A method can provide a plurality of switches, a plurality of hosts, the plurality of hosts being interconnected via the plurality of switches, wherein a host of the plurality of hosts comprises a multicast sender node, the sender node comprising a system image generation module and a current message sequence module. The method can organize the plurality of switches into two rails, the two or more rails providing redundant connectivity between the plurality of hosts. The method can send two or more duplicate multicast packets on different rails. Upon a receiving node receiving at least two versions of the same multicast packet, only one will be delivered to the communication stack/clients above the layer that handles the encapsulation header.