Network Fault Diagnosis via Temporal Metric Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault diagnosis systems in Self-Organizing Networks face challenges in accurately identifying the root cause of faults due to reliance on primitive methods, limited information availability, and the use of thresholds that ignore small variations and specific degraded patterns, leading to incorrect classifications and increased diagnosis errors.
Innovation Solution
An apparatus and method that utilize temporal analysis by correlating primary network metrics with indicator metrics, employing weighted correlation and shifting techniques to determine fault correlation values, and incorporating expert knowledge to improve diagnosis accuracy, particularly in heterogeneous networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If threshold-based anomaly detection is used to identify faults, then the system can detect obvious deviations from normal behavior, but it ignores small variations and specific degraded patterns leading to diagnosis errors
Solution Approach 1:
The patent changes the detection parameter from binary threshold-based anomaly detection to continuous correlation coefficient calculation. Instead of checking whether metrics violate predefined thresholds, the system calculates correlation coefficients between current and historical metric values, capturing subtle temporal patterns and small variations that thresholds would miss. This allows the system to detect degraded patterns while maintaining information about small variations.
2Reliability
If primitive fault diagnosis approaches are used, then the system is simple to implement, but it cannot accurately identify root causes in complex Self-Organizing Networks
Solution Approach 1:
The patent introduces correlation coefficients as an intermediary metric that bridges raw network performance data and fault root cause identification. Instead of directly mapping complex network metrics to fault causes, the system first computes correlation coefficients that capture temporal relationships, then uses these coefficients to identify root causes. This intermediary approach simplifies the diagnosis process while improving accuracy in complex Self-Organizing Networks.
Solution Approach 2:
The patent adds a temporal dimension to fault diagnosis by analyzing the evolution of correlation coefficients over time. Instead of examining static metric values, the system tracks how correlations between metrics change, enabling identification of root causes through temporal patterns. This dimensional transformation allows accurate diagnosis in dynamic Self-Organizing Networks without requiring overly complex systems.
3Productivity
If fixed thresholds are used for anomaly classification, then the decision process is simple and fast, but it leads to drastic decisions that ignore specific degraded patterns
Solution Approach 1:
The patent replaces static fixed thresholds with dynamic correlation coefficient calculations that adapt to the specific temporal patterns of each fault scenario. Instead of using predetermined threshold values, the system dynamically computes correlations between current and historical metrics, allowing the diagnosis process to recognize specific degraded patterns while maintaining fast processing speeds through efficient correlation calculations.
Data Source
AI summary
An apparatus and corresponding method for determining a cause of a fault in a network. The apparatus comprises a means for receiving, which may be a receiver, configured to receive a plurality of time separated samples of a primary network metric that are indicative of a fault in the network, and a plurality of time separated samples of one or more indicator network metrics. The apparatus comprises a means for correlating, which may be a correlator, configured to determine one or more metric correlation values relating to dependences between the samples of the primary network metric and the samples of each of the one or more indicator network metrics. The correlator is further configured to determine one or more fault correlation values relating to dependences between the one or more metric correlation values and a plurality of stored metric correlation values associated with a fault cause. The apparatus comprises a means for fault determining, which may be a fault determiner, configured to determine a cause of the fault based on the one or more fault correlation values.