Operation Management Apparatus for Fault Detection via Correlation Collapse
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional operation management apparatuses for information-telecommunications systems face challenges in accurately detecting and localizing performance-degrading faults, particularly in systems with maldistributed elements, as they often fail to detect faults in limited elements and are prone to erroneous fault detection due to varying performance information thresholds.
Innovation Solution
An operation management apparatus that monitors performance information, extracts correlation variants, and accumulates normal and fault ranges to discriminate fault models, allowing for accurate fault detection and localization by determining whether the correlation collapse range falls within predefined normal or fault ranges, using programs stored in digital storage media.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If thresholds are set to detect faults in performance information, then fault detection capability is improved, but erroneous detection increases when performance information varies greatly
Solution Approach 1:
The patent changes the detection parameter from absolute threshold values to correlation coefficients. Instead of comparing performance information against fixed thresholds, the system calculates correlation coefficients between multiple performance parameters to detect faults. This resolves the contradiction by making detection insensitive to absolute performance variations while maintaining sensitivity to abnormal correlations.
Solution Approach 2:
The patent introduces correlation coefficients as an intermediary metric between raw performance information and fault detection. Rather than directly comparing performance values to thresholds, the system uses correlation coefficients to mediate the detection process, enabling accurate fault identification without false alarms caused by performance variations.
2Reliability
If correlation model is produced to detect faults based on correlation collapse frequency, then system-wide fault detection is improved, but faults in limited elements with maldistributed performance are not detected
Solution Approach 1:
The patent segments the fault detection process into two levels: system-wide correlation analysis and local element correlation analysis. By dividing the monitoring scope into different granularity levels, the system can simultaneously detect both system-wide faults and localized element faults, resolving the contradiction between comprehensive coverage and precise local detection.
Solution Approach 2:
The patent adds a new dimension to fault detection by analyzing correlation relationships between multiple performance parameters simultaneously. Instead of monitoring single parameters in isolation, the system examines multi-dimensional correlation patterns, enabling detection of faults that manifest as correlation changes rather than absolute value changes.
3Reliability
If thresholds are set to detect significant faults, then stable system operation detection is improved, but performance degradation faults become difficult to detect
Solution Approach 1:
The patent changes the detection approach from monitoring absolute performance values against thresholds to monitoring correlation relationships between parameters. This enables detection of performance degradation through changes in correlation patterns, even when absolute values remain within acceptable ranges, thus detecting both stable operation and degradation without missing subtle faults.
Data Source
AI summary
An operation management apparatus monitors the performance information of a system executing services and extracts correlation variants of the performance information, thus allowing the management to detect and localize performance-degrading faults of the system. It is determined whether or not the range of correlation collapse ascribed to the performance information falls within the normal range of correlation collapse measured in the normal operation of the system. A history is retained with regard to the range of correlation collapse ascribed to the performance information which does not fall within the normal range of correlation collapse. It is determined whether or not the history accumulating the predetermined number of ranges of correlation collapse approximates to the fault range of correlation collapse measured in the faulty operation of the system, thus discriminating a fault model on the basis of the performance information.