Forwarding Device Service Information Extraction for Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fault detection models in communication networks require mirroring and sending entire service flows, leading to high network resource consumption.
Innovation Solution
A method where a forwarding device extracts service information, including identification information and key performance indicators (KPIs), from service flows and sends this reduced data to a first device for training a fault detection model, reducing network resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the access device or forwarding device mirrors the service flow and sends it to the analysis platform for fault detection model training, then the fault detection model can be obtained through training based on the service flow of the network object, but a large quantity of network resources is consumed
Solution Approach 1:
The patent extracts only the essential features from the complete service flow for model training. Specifically, it extracts flow identification information, source and destination addresses, protocol types, and key performance indicators (KPIs) such as packet length, inter-arrival time, and sequence numbers. This extraction approach maintains fault detection capability while dramatically reducing the data volume that needs to be transmitted over the network, thereby resolving the contradiction between reliable fault detection and network resource consumption.
Solution Approach 2:
The patent applies local quality by differentiating the data processing approach between training phase and detection phase. During training, the system processes extracted feature data rather than complete service flows. During detection, the system uses the trained model to evaluate new service flows based on the same extracted features. This localized optimization of data quality and quantity at different stages resolves the contradiction by ensuring sufficient data for training while minimizing network resource usage during operation.
2Loss of information
If the service flow is mirrored and transmitted to the analysis platform, then comprehensive data for model training is available, but the data volume is large and consumes significant network bandwidth
Solution Approach 1:
The patent implements feature extraction that identifies and removes redundant information from the service flow. It extracts only the critical elements needed for fault detection: flow identification, address information, protocol types, and KPIs. This extraction maintains the essential information content required for comprehensive model training while reducing the transmitted data volume by a significant margin, thus resolving the contradiction between data completeness and transmission quantity.
Solution Approach 2:
The patent transforms the service flow data from its original comprehensive form into a simplified feature representation. It changes the parameters by selecting specific measurable attributes (KPIs) such as packet inter-arrival time, packet length distribution, and sequence number patterns, rather than transmitting the complete raw service flow. This parameter transformation preserves the informative content needed for training while dramatically reducing the data volume to be transmitted.
Data Source
AI summary
A forwarding device receives at least one service flow; the forwarding device obtains service information of the at least one service flow, where the service information of the service flow includes identification information of a network object to which the service flow belongs and M key performance indicators KPIs of the service flow, M is an integer greater than 0, and the network object includes one or more devices; and the forwarding device sends training information to a first device, where the training information includes the service information of the at least one service flow or a feature set obtained based on the service information of the at least one service flow, the training information is used to train a fault detection model, and the fault detection model is used to detect whether the network object is in a faulty state.