Regression Model Recalculation for Disk Utilization Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional performance management methods struggle with predicting disk utilization rates accurately across different usage patterns in systems, leading to increased prediction errors and reduced accuracy.
Innovation Solution
A monitoring data analyzing apparatus and method that combines regression models generated for each usage pattern group, using log data to recalculate models based on characteristic data items extracted from test target log data, improving prediction accuracy by adapting to specific usage patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single regression model is generated using monitoring data from all usage patterns, then the model can be applied universally to any pattern, but prediction accuracy deteriorates because utilization rates differ substantially across patterns
Solution Approach 1:
The patent segments the monitoring data into multiple groups based on usage patterns (e.g., daytime browsing vs. nighttime batch processing). A separate regression model is generated for each segment, allowing each model to capture the specific characteristics of its usage pattern. This segmentation resolves the contradiction by sacrificing universal applicability in favor of pattern-specific prediction accuracy.
Solution Approach 2:
The patent implements a dynamic model selection mechanism that automatically identifies the current usage pattern and selects the appropriate regression model. The system transitions between different models based on real-time usage conditions, making the prediction system adaptive rather than static. This dynamic approach maintains high prediction accuracy across varying usage patterns while preserving a form of universal applicability through automated model switching.
2Measurement precision
If regression models are generated separately for each usage pattern group, then prediction accuracy improves for specific patterns, but system complexity increases due to multiple models
Solution Approach 1:
The patent creates a universal prediction system that handles multiple usage patterns through a single integrated framework. The system uses a usage pattern identification module that can recognize different patterns and route to appropriate models, making the overall system multi-functional without requiring separate independent prediction systems for each pattern. This universality manages complexity by providing a unified interface and automated selection process.
Solution Approach 2:
The patent introduces an intermediary usage pattern identification module that acts as a mediator between the monitoring data and the multiple regression models. This intermediary automatically determines which usage pattern is currently active and selects the corresponding model, thereby managing the complexity of multiple models through a single decision-making layer rather than requiring direct management of each model's applicability.
3Measurement precision
If the regression model is recalculated frequently to adapt to changing usage patterns, then prediction accuracy is maintained, but computational overhead and processing time increase
Solution Approach 1:
The patent performs preliminary actions by pre-segmenting the monitoring data into usage pattern groups and pre-generating regression models for each segment during system setup or off-peak periods. This preliminary preparation allows the system to quickly switch between pre-computed models during operation without performing complex recalculations in real-time, thereby maintaining prediction accuracy while minimizing computational overhead during active use.
Solution Approach 2:
The patent implements periodic model recalculation rather than continuous recalculation. The system recalculates or updates regression models at predetermined intervals or when significant changes in usage patterns are detected, rather than continuously adapting to every data point. This periodic approach maintains prediction accuracy over time while significantly reducing computational overhead compared to continuous recalculation.
Data Source
AI summary
An object of the present invention is to improve prediction accuracy for a performance value of a monitoring target system even if the performance value is substantially different for each of patterns of use. A monitoring data analyzing apparatus includes a regression-model recalculating section 14 configured to combine regression models, which are generated for each of a plurality of groups into which log data including monitoring data in a monitoring target system set as a target of performance management is classified, using the log data belonging to the groups corresponding to the regression models and test target log data, which is the log data set as a target of a performance test, to recalculate the regression models.