Causal Graph Root Cause Analysis for Cloud System Anomalies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for root cause analysis of computerized system anomalies, particularly in cloud computing, face challenges in explaining model predictions and identifying true causes of anomalies due to inefficiency and scalability issues with complex datasets, and fail to effectively drill down from symptom detection to causal inference.
Innovation Solution
A computer-implemented method that monitors key performance indicators (KPIs) to detect anomalies, determines a troubleshooting time window, and uses a combination of feature importance attribution and causality algorithms to identify a strict subset of KPIs, constructing a causal graph that links abnormal KPIs to potential explanatory KPIs, enabling experts to investigate causality chains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If baseline causality methods are used to analyze large datasets, then causal relationships can be identified, but the methods suffer from inefficiency and scalability issues
Solution Approach 1:
The patent segments the causal analysis process into multiple stages: first identifying abnormal KPIs using anomaly detection, then selecting explanatory KPIs based on feature importance, and finally constructing causal graphs using a reduced subset of KPIs. This segmentation allows the system to handle large datasets efficiently by processing information in manageable chunks rather than analyzing all KPIs simultaneously.
Solution Approach 2:
The patent extracts only the necessary KPIs for causal analysis by first identifying abnormal KPIs and then selecting explanatory KPIs based on feature importance scores. This extraction approach reduces the dataset to a manageable subset containing only the KPIs relevant to the anomaly investigation, thereby improving analysis efficiency while maintaining causal relationship identification accuracy.
2Loss of information
If explainability methods are used to analyze anomalies, then model predictions can be explained, but the methods fail to drill in deeper from causal inference of symptoms to root cause analysis
Solution Approach 1:
The patent performs preliminary actions by first identifying abnormal KPIs and selecting explanatory KPIs before constructing the causal graph. This preliminary filtering ensures that the causal analysis focuses only on the most relevant KPIs, enabling deeper root cause analysis without overwhelming complexity.
Solution Approach 2:
The patent introduces an intermediary layer of feature importance scoring that bridges anomaly detection and causal graph construction. This intermediary mechanism selects explanatory KPIs based on their relevance to abnormal KPIs, enabling the system to drill down from symptom detection to root cause analysis in a structured manner.
3Reliability
If a large subset of KPIs is used for causal analysis, then more comprehensive causal relationships can be identified, but the computational complexity increases significantly
Solution Approach 1:
The patent applies partial action by analyzing only the subset of KPIs that are both abnormal and have high feature importance scores. This partial analysis approach provides sufficient causal insights for troubleshooting while significantly reducing computational complexity compared to analyzing all KPIs in the system.
Data Source
AI summary
An embodiment for root cause analysis of computerized system anomalies is provided. The embodiment may include monitoring key performance indicators (KPIs) for a computerized system, wherein KPI values of the monitored KPIs form respective timeseries. The embodiment may include detecting an anomaly in the computerized system based on the monitored KPIs. The embodiment may include determining a troubleshooting time window extending over a given time period. The embodiment may include identifying a strict subset of the monitored KPIs based on portions of the respective timeseries spanning the given time period. The strict subset comprises abnormal KPIs (aKPIs) and potential explanatory KPIs (xKPIs). The embodiment may include obtaining a causal graph of vertices mapping KPIs of the strict subset by running a causality algorithm to evaluate weights of directed edges connecting the vertices and accordingly obtain one or more directed paths. The embodiment may include returning the obtained causal graph.