ML Contention Detection for Workload Groups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In enterprise-class computing systems, identifying the cause of performance anomalies due to workload contention across different workload groups is challenging, as existing methods lack effective automated solutions, leading to resource inefficiencies and manual, time-consuming analysis.
Innovation Solution
A computer-implemented method and system utilizing machine-learning models to separate contention-related data from system performance data, predicting contention instances and impact values, and applying a multi-stage approach with Gradient Boosted Tree models and SHAP values to identify root causes of performance bottlenecks across workload groups.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual analysis methods are used to identify contention causes, then analysis accuracy can be maintained, but time consumption and manual effort increase significantly
Solution Approach 1:
The patent replaces manual mechanical analysis with automated machine learning systems. Multiple ML models (random forest, gradient boosting, neural networks) automatically analyze system performance data to identify contention causes, substituting human analysts while maintaining high accuracy through ensemble methods and feature importance analysis.
Solution Approach 2:
The patent introduces machine learning models as intermediary systems between raw performance data and contention cause identification. These models process large volumes of system data, extract relevant features, and provide structured insights that would be difficult for manual analysis to obtain efficiently.
2Measurement precision
If comprehensive system data is analyzed to identify contention causes, then detection accuracy improves, but system complexity and computational requirements increase
Solution Approach 1:
The patent segments the complex analysis task into multiple independent ML models, each specializing in different aspects of contention detection. The system divides performance data into separate feature sets and applies different algorithms (random forest for categorical features, gradient boosting for numerical features, neural networks for pattern recognition), reducing the complexity of any single model while maintaining comprehensive analysis capability.
Solution Approach 2:
The patent employs multiple ML models beyond what a single model would provide, using ensemble methods to achieve higher accuracy. By applying several models with different strengths to the same data, the system achieves robust contention detection that compensates for individual model limitations.
Data Source
AI summary
A computer-implemented method for identifying a cause of a performance anomaly of a computer system executing workloads in different workload groups is disclosed. The method comprises receiving system performance data, separating contention-related data and non-contention related data within the received system management data, feeding a first part of the contention-related data to a first machine-learning system comprising a trained first machine-learning model for predicting first contention instances and related first impact values as output, and feeding a second part of the contention-related data scaled with the first impact values to a second trained machine-learning system comprising a trained second machine-learning model for predicting second contention instances and related second impact values for the different workload groups as output.