Automated Root Cause Analysis for Network Service Degradation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solutions for root cause analysis in communication networks are limited by their reliance on decision trees, which require a training period, may not accurately reflect real-time network status, and often miss causes due to prior decisions, leading to delayed and impractical identification of service degradation issues, especially in heterogeneous environments with limited automation and interoperability.
Innovation Solution
A method and apparatus for automated root cause analysis using simple decision logic applied to network resource measurements, identifying underperforming resources and their dependents, and outputting causes of service quality degradation, which employs a top-down approach with deterministic rules and algorithms to analyze accessibility, retainability, and integrity performance metrics in real-time, without the need for domain-specific knowledge or human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If decision trees are used for root cause analysis, then automation is provided, but the system requires a training period and may not accurately reflect real-time network status
Solution Approach 1:
The patent extracts the decision logic from complex trained decision trees and implements it as simple deterministic rules that can be immediately applied to measurements. The system takes out the need for training data and probability calculations, using instead straightforward if-then logic that provides immediate root cause identification without delay.
Solution Approach 2:
The patent replaces complex, expensive decision tree models that require extensive training with simple, lightweight decision rules that can be deployed immediately. These simple rules act as disposable alternatives to the heavy training infrastructure, providing fast results without the overhead of maintaining complex trained models.
2Ease of operation
If decision trees with probabilities are used, then branching decisions can be made, but causes may be missed due to earlier decisions and probabilities may not be accurate
Solution Approach 1:
The patent segments the root cause analysis into independent evaluation of each potential cause rather than forcing sequential branching decisions. Each cause is evaluated independently against the measurements, allowing multiple causes to be identified simultaneously without one decision preventing the detection of others.
Solution Approach 2:
Instead of using probabilities to guide which branch to follow (top-down with bias), the patent inverts the approach by evaluating all possible causes and selecting those that best explain the measurements. This bottom-up inversion ensures that no potential cause is prematurely eliminated by probability-based filtering.
3Measurement precision
If complex systems and algorithms are used for analysis, then specific service problems can be found, but the system is limited to specific problems and requires huge human intervention
Solution Approach 1:
The patent creates a universal decision logic framework that can handle multiple types of service problems through the same simple rules. Instead of having separate complex algorithms for each service type, a single multi-functional decision logic evaluates measurements against resource dependencies to identify root causes across diverse services and problems.
Solution Approach 2:
The system enables self-service automation where the simple decision logic automatically correlates measurements with resource dependencies without requiring human domain knowledge. The system serves itself by using the measurements and predefined resource models to identify root causes, eliminating the need for human experts to configure and maintain complex service-specific algorithms.
4Measurement precision
If point solutions with different performance metrics are used, then specific measurements can be provided, but interoperability and compatibility are poor
Solution Approach 1:
The patent merges measurements from multiple different network resources and performance metrics into a unified analysis framework. The simple decision logic correlates measurements across heterogeneous resources by evaluating them against resource dependencies, combining information from probe-based solutions, network monitors, and other sources into a coherent root cause identification.
Data Source
AI summary
Causes of service quality degradation in a communication network are determined for automated root cause analysis by identifying occurrence of degradation in service quality within a user session; identifying an underperforming resource of the communication network during occurrence of said identified degradation in service quality; in response to determining that the underperforming resource has at least one dependant resource, identifying any underperforming dependant resources during occurrence of said identified degradation in service quality; and outputting a plurality of causes of the service quality degradation including the identity of each of the identified underperforming resources.