Graph Neural Network Virtual Network Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network management technologies, particularly in virtualized environments, face challenges in real-time optimization due to reliance on human judgment and costly operational personnel, with existing machine learning methods struggling to effectively learn and apply optimal virtual network function (VNF) management policies efficiently.
Innovation Solution
The implementation of a graph neural network (GNN) based method for virtual network management, which preprocesses network data, converts it into a graph format, and uses a feedforward neural network (FNN) to learn node state information, allowing for efficient VNF management decisions such as adding, removing, or optimizing VNF instances in real-time, while addressing class imbalance through weight adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Integer Linear Programming (ILP) is used to optimize network management, then optimal network management solution is obtained, but it takes a relatively long time to find the optimal solution making it unsuitable for real-time network management
Solution Approach 1:
The patent creates a graph neural network model that copies and learns from historical network management data and ILP solutions. Instead of solving ILP problems in real-time, the system trains a GNN model offline using ILP-generated ground truth data, then deploys the trained model for rapid real-time predictions, effectively copying the optimal decision-making capability without the computational burden of repeated ILP solving
Solution Approach 2:
The system performs preliminary action by pre-training the graph neural network model offline using historical network data and ILP-generated optimal solutions. This offline training phase prepares the model in advance so that during real-time network management, the pre-trained model can provide rapid predictions without requiring time-consuming ILP computations at runtime
2Loss of time
If machine learning technology is applied to network management, then network optimization time is reduced, but existing ML methods struggle to effectively learn and apply optimal VNF management policies in virtualized environments
Solution Approach 1:
The patent transforms network management data into graph-structured parameters with specific node and edge features tailored to virtualized network environments. Node features include VNF states, resource allocations, and performance metrics, while edge features capture network connectivity and traffic patterns. This parameter transformation enables the GNN to effectively learn complex VNF management policies by representing network states in a format that preserves structural relationships
Solution Approach 2:
The graph neural network model serves as an intermediary between raw network data and VNF management decisions. The GNN processes complex network states represented as graphs, learns optimal management policies through training, and outputs actionable VNF deployment decisions. This intermediary model bridges the gap between raw data and effective policy application in virtualized network environments
3Measurement precision
If human judgment and specialized operational personnel are used for network management, then complex network management decisions are made accurately, but it requires costly operational personnel and becomes more difficult as network complexity increases
Solution Approach 1:
The system implements self-service by enabling automated network management through the graph neural network model. The GNN automatically learns optimal VNF management policies from historical data and makes real-time management decisions without requiring human operational personnel. The system serves itself by continuously learning from network data and autonomously making deployment decisions, eliminating dependence on specialized human expertise
Data Source
AI summary
A method for graph neural network-based virtual network management may comprise: preprocessing collected network data; converting node and edge data among the data for each category generated in the preprocessing step into a graph; learning the network data expressed in a matrix generated in the step of converting into the graph using a graph neural network (GNN); and learning node state information for each node generated through the learning using the GNN using a feedforward neural network (FNN) together with service list data.