Hierarchical Problem Classification for IT Root Cause Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing problem determination and resolution (PDR) systems are limited in their ability to identify the root cause of issues in complex multi-tier IT environments, often focusing on specific types of problems and ignoring relevant information, leading to inefficient and labor-intensive processes that increase operational costs.
Innovation Solution
A system and method that utilize a monitor and data collector to gather data from the IT environment, compare it to historic patterns using a hierarchical structure with trained linear classifiers, and automatically generate the root cause of problems, allowing for efficient and accurate identification and resolution of issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing PDR systems use problem-specific approaches to determine root causes, then they can accurately identify specific types of problems (network, hardware, storage, database, web), but they lack versatility and cannot effectively search for other types of problems in complex multi-tier IT environments
Solution Approach 1:
The patent creates a universal PDR system that can handle multiple types of problems (network, hardware, storage, database, web, and other emerging problem types) through a single unified framework. The system uses a standardized data collection mechanism that gathers relevant information across all problem types, and a classification algorithm that automatically determines the problem category and root cause without requiring separate specialized systems for each problem type.
Solution Approach 2:
The patent segments the complex problem determination process into distinct functional components: data collection from multiple sources, data preprocessing and normalization, problem classification, root cause identification, and solution recommendation. This segmentation allows each component to be optimized independently while maintaining overall system versatility and accuracy.
2Measurement precision
If prior art methodologies use only one type of available information (logs or performance metrics), then they simplify the analysis process, but they ignore relevant information that may be critical for accurate problem determination
Solution Approach 1:
The patent merges multiple information sources (system logs, performance metrics, error messages, configuration data, and event streams) into a unified analysis framework. The system collects and integrates data from diverse sources simultaneously, allowing the classification algorithm to consider all relevant information together to improve detection accuracy while managing complexity through standardized integration procedures.
Solution Approach 2:
The patent introduces intermediary components including data normalization layers and feature extraction modules that bridge different information sources. These intermediaries standardize diverse data formats and extract relevant features, making it easier to process multiple information types without overwhelming system complexity.
3Measurement precision
If semi-supervised approaches using human testers for annotation judgment are used for problem classification, then classification accuracy can be maintained, but the process becomes labor intensive and costly
Solution Approach 1:
The patent implements a self-service classification system where the algorithm automatically learns from historical problem data and performs classification without requiring human testers for each new problem. The system uses supervised learning during an initial training phase with annotated historical data, then autonomously classifies new problems by comparing them against learned patterns, dramatically increasing throughput while maintaining accuracy.
Solution Approach 2:
The patent incorporates feedback mechanisms where classification results are continuously evaluated and used to refine the model. When classification outcomes are validated (either automatically or through minimal human feedback), the system learns from these results and improves future classifications, maintaining high accuracy while reducing the need for continuous human intervention.
4Adaptability or versatility
If automated classification approaches use a flat structure of taxonomy to classify problem instances, then the classification process can be simplified, but scalability is limited when the problem space becomes large as in complex modern IT environments
Solution Approach 1:
The patent transitions from a flat, single-level taxonomy to a hierarchical, multi-dimensional classification structure. Problems are classified across multiple dimensions and levels (e.g., problem type, severity, component, root cause category), allowing the system to scale to complex IT environments by organizing the vast problem space into manageable hierarchical layers rather than a single flat category list.
Data Source
AI summary
A system and method of problem determination and resolution utilizes enhanced problem classification, and effectively categorizes any problem a user experiences by leveraging all available data to recognize the specific problem. Historical problem data is labeled with the cause of that problem and is analyzed to learn problem patterns. The historical problem data is classified into a predefined hierarchical structure of taxonomies by using an incremental online learning algorithm. The hierarchical structure and learned patterns are utilized to recognize problems and generate the root cause of the problem when given a new set of monitoring data and log data.