Network Load Sharing via Multi-Node Hash Gene Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional routing technologies fail to ensure all links in a network are non-congested, leading to load imbalance and inefficient packet forwarding.
Innovation Solution
A method where a network device selects a packet flow and determines non-congested links across multiple forwarding nodes by using a hash gene, ensuring the packet flow is forwarded through at least two non-congested links, and saves the hash gene in the source node to maintain balanced network load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional routing technology is used, then packet forwarding is simple, but load imbalance occurs and some links remain congested while others are idle
Solution Approach 1:
The controller performs preliminary actions by calculating hash genes for multiple forwarding nodes in advance, determining non-congested links before packet flow occurs. This allows the source node to select appropriate hash genes that ensure packet flows will traverse non-congested links across the network, preventing congestion before it happens.
Solution Approach 2:
The controller acts as an intermediary between the source node and forwarding nodes. It calculates hash genes and provides them to the source node, which then uses these hash genes to select non-congested links for packet forwarding. This intermediary mechanism enables coordinated load balancing across multiple forwarding nodes without requiring complex direct communication between them.
2Productivity
If ECMP technology is used at source node, then local load balancing is achieved, but load imbalance cannot be resolved at other forwarding nodes
Solution Approach 1:
The invention extends load balancing from a single-dimension approach (source node only) to a multi-dimensional approach by considering multiple forwarding nodes simultaneously. The controller calculates hash genes for each forwarding node's perspective, ensuring that packet flows are distributed across non-congested links at multiple dimensions of the network, not just at the source node.
Solution Approach 2:
The controller uses feedback information about link congestion status at multiple forwarding nodes to adjust hash gene selection. By monitoring which links are congested at which forwarding nodes and incorporating this feedback into hash gene calculation, the system dynamically adapts to prevent congestion propagation throughout the network.
3Speed
If hash gene is selected without verifying downstream links, then forwarding speed is fast, but congestion may occur at downstream forwarding nodes
Solution Approach 1:
The controller performs preliminary verification of downstream links by calculating what hash gene would be selected at each forwarding node and checking if the corresponding links are congested. This advance verification ensures that only hash genes leading to non-congested paths are provided to the source node, preventing downstream congestion before packet flow begins.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Embodiments of the present invention disclose a method for implementing load sharing. The method includes: for a congested first link on a first forwarding node, selecting, by a network device, a packet flow forwarded by using the first link; selecting, by the network device, a second link that may be used to forward the packet flow and that is not congested after available bandwidth of the second link is occupied by the packet flow, where the second link is a link between the first forwarding node and a second forwarding node; selecting, by the network device, a first hash gene corresponding to the second link; determining, by the network device, that a third link is not in a congested state after available bandwidth of the third link is occupied by the packet flow; and saving the first hash gene in a source node of the packet flow, where the third link is a link that is on the second forwarding node and that is corresponding to the first hash gene, and the third link is a link between the second forwarding node and a third forwarding node. In addition, the embodiments of the present invention further disclose an apparatus for implementing load sharing.