Error Dynamics Analysis via Association Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current analytics tools for analyzing error logs in cloud systems are unable to determine the dynamics of errors, leading to false alarms and inefficiencies in identifying problematic code segments.
Innovation Solution
A method involving the generation of association rules models from error logs to identify patterns and relationships between error messages, computing components, and tasks, using templatized error messages and graph databases to analyze error dynamics across different instances and time periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current analytics tools are used to analyze error logs, then basic error detection is possible, but the ability to determine error dynamics and identify patterns is insufficient
Solution Approach 1:
The patent segments the error log analysis process into distinct components: error message extraction, association rule generation, pattern identification, and dynamics analysis. This segmentation allows each component to be optimized independently, improving measurement precision for error dynamics while managing overall system complexity through modular architecture.
Solution Approach 2:
The patent introduces association rules as an intermediary mechanism between raw error logs and pattern identification. These rules serve as a mediator that transforms unstructured error messages into structured relationships, enabling dynamics analysis without requiring direct complex processing of raw logs.
2Reliability
If detailed analysis of all error logs is performed, then comprehensive error understanding is achieved, but false alarms increase
Solution Approach 1:
The patent implements feedback mechanisms where association rules are continuously refined based on pattern identification results. The system learns from analyzed errors and adjusts rule sensitivity, providing feedback that reduces false alarms while maintaining comprehensive error understanding through iterative improvement of analysis accuracy.
Solution Approach 2:
The patent changes parameters of error analysis by transforming static error messages into dynamic association rules with configurable confidence thresholds and support values. This parameter transformation allows the system to adjust analysis sensitivity, reducing false alarms while maintaining reliable error detection through optimized parameter settings.
3Productivity
If traditional log analysis methods are used, then simple error counting is possible, but identification of problematic code segments is inefficient
Solution Approach 1:
The patent adds a new dimension to log analysis by introducing temporal and relational dimensions through association rules. Instead of simple error counting, the system analyzes error sequences, relationships between errors and code segments, and patterns across multiple dimensions, dramatically improving code review efficiency while reducing time to identify problematic segments through multi-dimensional pattern matching.
Data Source
AI summary
A method, a system, and a computer program product for analyzing error messages. A first error log generated as a result of an execution of at least one task of a computing system at a first instance is received. The first error log include a plurality of first error messages. A first association rules model is generated using the first error messages. The first association rules model includes a plurality of association rules defining one or more relationships. A second error log, including a plurality of second error messages, generated as a result of an execution of the task at a second instance is received and a second association rules model is generated using the second error messages. Based on the first and second association rules models, at least one error message pattern associated with execution of the at least one task is determined.