Per-Hop Seeding for Hash-Based Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hash-based load balancing in large multi-hop, multi-path networks leads to poor distribution of traffic across aggregate members, causing node starvation and imbalance, as the same mathematical transformation is applied at each node, resulting in correlated next-hop/link selection and disproportionate traffic distribution.
Innovation Solution
Implementing per-hop seeding techniques, including seed expansion, seed manipulation, dynamic hash configuration modes, and member selection, to randomize flow distribution at each node, reducing correlation in next-hop/link selection and minimizing traffic imbalances by allowing unique mathematical transformations and configurations at each node.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the same mathematical transformation is applied at each node for hash-based load balancing, then packet order is preserved and flow binding is deterministic, but traffic distribution across aggregate members becomes poor and node starvation occurs
Solution Approach 1:
The patent applies dynamics by making the hash transformation configurable and variable at each node rather than fixed. Each node can be configured with different transformation parameters, allowing the system to adapt to different network conditions and achieve better load distribution while maintaining determinism through controlled configuration.
Solution Approach 2:
The patent implements local quality by allowing each node to have its own specific hash transformation configuration. Instead of using a uniform transformation across all nodes, each node can be configured with local parameters that optimize traffic distribution for that specific node's position in the network hierarchy.
2Loss of information
If deterministic hash-based load balancing is used to distribute flows to aggregate members, then packet flow can be traced, but traffic distribution becomes correlated and imbalanced in multi-hop networks
Solution Approach 1:
The patent implements local quality by allowing each node to have its own specific hash transformation configuration. Instead of using a uniform transformation across all nodes, each node can be configured with local parameters that optimize traffic distribution for that specific node's position in the network hierarchy.
Solution Approach 2:
The patent applies dynamics by making the hash transformation configurable and variable at each node rather than fixed. Each node can be configured with different transformation parameters, allowing the system to adapt to different network conditions and achieve better load distribution while maintaining determinism through controlled configuration.
3Stability of the object's composition
If fixed flow binding to next hop is used, then packet sequence is guaranteed, but nodes far from the root node experience starvation
Solution Approach 1:
The patent implements local quality by allowing each node to have its own specific hash transformation configuration. Instead of using a uniform transformation across all nodes, each node can be configured with local parameters that optimize traffic distribution for that specific node's position in the network hierarchy.
Solution Approach 2:
The patent applies dynamics by making the hash transformation configurable and variable at each node rather than fixed. Each node can be configured with different transformation parameters, allowing the system to adapt to different network conditions and achieve better load distribution while maintaining determinism through controlled configuration.
Data Source
AI summary
Methods and apparatus for improving hash-based load balancing with per-hop seeding are disclosed. The methods and apparatus described herein provide a set of techniques that enable nodes to perform differing mathematical transformations when selecting a destination link. The techniques include manipulation of seeds, hash configuration mode randomization at a per node basis, per node/microflow basis or per microflow basis, seed index generation, and member selection. A node can utilize any, or all, of the techniques presented in this disclosure simultaneously to improve traffic distribution and avoid path starvation with a degree of determinism.