Multi-Path Load Balancing by Traffic Type and Network State

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing network routing technologies use a one-size-fits-all approach for load balancing across multi-path groups, leading to suboptimal performance for different types of network traffic such as TCP and RDMA, as they do not account for the specific sensitivity of each traffic type to packet reordering and latency.

Innovation Solution

Adaptive selection of load balancing algorithms based on underlying network applications, using different algorithms for TCP and RDMA traffic to maintain packet ordering and optimize path utilization respectively.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If a one-size-fits-all load balancing algorithm is used for all traffic types, then device complexity is reduced, but network performance deteriorates due to suboptimal packet ordering and latency for different traffic types

Engineering Contradiction:
Improveload balancing configuration complexityVSAvoidnetwork performance
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent applies local quality by differentiating load balancing algorithms based on traffic type. TCP traffic uses an algorithm that maintains packet ordering to ensure reliable delivery, while RDMA traffic uses an algorithm that optimizes for low latency and allows reordering. This localized differentiation of handling methods resolves the contradiction by improving overall network performance without requiring complete complexity increase across all traffic types.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system dynamically selects load balancing algorithms based on the detected traffic type. The controller monitors packet characteristics and switches between ordering-sensitive algorithms for TCP and reordering-tolerant algorithms for RDMA. This dynamic adaptation allows the system to optimize performance for each traffic type while maintaining manageable complexity through automated decision-making.

Inventive Principle:
Principle #15Dynamics

2Reliability

If packet ordering is maintained for all traffic, then TCP reliability is improved, but RDMA throughput deteriorates due to unnecessary retransmissions from expected ordering

Engineering Contradiction:
ImproveTCP packet delivery reliabilityVSAvoidRDMA throughput
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent implements local quality by applying different packet handling policies to different traffic types. For TCP traffic, the system maintains strict packet ordering to ensure reliable delivery and prevent retransmissions. For RDMA traffic, the system allows packet reordering and uses algorithms that optimize for throughput, knowing that RDMA protocols can handle out-of-order packets without retransmission penalties. This differentiated approach resolves the contradiction by optimizing each traffic type according to its specific requirements.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system changes the load balancing algorithm parameter based on traffic type detection. When TCP traffic is detected, the system selects an algorithm with ordering preservation enabled. When RDMA traffic is detected, the system switches to an algorithm with reordering tolerance enabled. This parameter change allows the same infrastructure to optimize for different performance characteristics without manual intervention.

Inventive Principle:
Principle #35Parameter changes

3Device complexity

If a single load balancing algorithm is used for all paths, then device complexity is reduced, but path utilization efficiency deteriorates for different traffic types

Engineering Contradiction:
Improvepath selection complexityVSAvoidpath utilization efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent applies local quality by selecting different path selection algorithms based on the underlying traffic type. For TCP traffic, the system uses an algorithm that distributes packets across multiple paths while maintaining ordering, optimizing for reliability. For RDMA traffic, the system uses an algorithm that aggressively utilizes all available paths with reordering tolerance, optimizing for throughput. This localized optimization improves overall path utilization efficiency without requiring complex manual configuration.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The load balancing system performs self-service by automatically detecting traffic types and selecting appropriate algorithms without external intervention. The controller monitors packet characteristics and autonomously switches between ordering-sensitive and reordering-tolerant algorithms. This self-service capability improves path utilization efficiency while keeping operational complexity manageable through automation.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS12500835B1Optimized path selection for multi-path groups
Publication Date: 2025.12.16 MARVELL ASIA PTE LTD
  • US12500835B1 patent drawing
  • US12500835B1 patent drawing
  • US12500835B1 patent drawing

AI summary

A packet to be forwarded over a computer network to a destination is received. A group of multiple network paths is available to forward to the packet to the destination. One or more path selection factors are determined to be used to identify a specific network load balancing algorithm to select a specific network path from the group of multiple network paths. The one or more path selection factors include at least one path selection factor determined based at least in part on a dynamic state of the computer network or a network node in the computer network. In response to selecting, by the specific network load balancing algorithm, the specific network path from among the group of multiple network paths, the packet is forwarded over the specific network path.