Partial Correlation Calculation for Multi-Tier Call Relationship Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining call relationships between processing types in multi-tier systems often inaccurately attribute response delays due to the lack of consideration for effects from all tiers, leading to erroneous determinations of call relationships.
Innovation Solution
A correlation coefficient calculation method that calculates and adjusts for the effects of total requests across multiple tiers, using partial correlation coefficients to accurately determine call relationships by removing the impact of tier-wide trends and simultaneous calls.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional correlation analysis is performed on time series data of processing types without removing tier-wide effects, then the calculation process is simple, but the measurement precision of call relationship determination deteriorates due to erroneous attributions
Solution Approach 1:
The patent segments the correlation analysis into multiple stages: first calculating tier-wide total request correlations, then calculating individual processing type correlations, and finally computing partial correlation coefficients that remove tier-wide effects. This segmentation allows precise isolation of direct call relationships from coincidental correlations.
Solution Approach 2:
The patent introduces total request time series data as an intermediary variable to mediate between individual processing type data and final call relationship determination. By using total requests as a mediator, the method removes confounding tier-wide effects before determining direct call relationships.
2Measurement precision
If partial correlation coefficients are calculated to remove tier-wide effects, then the measurement precision improves, but the calculation time increases
Solution Approach 1:
The patent performs preliminary calculations of total request time series and tier-wide correlation coefficients before calculating individual processing type correlations. By preparing these intermediary results in advance, the method avoids redundant calculations and reduces overall computation time despite the increased precision requirements.
3Measurement precision
If correlation analysis is performed on all processing types across tiers, then the detection completeness improves, but the difficulty of detecting and measuring genuine call relationships worsens due to false positives
Solution Approach 1:
The patent extracts and removes tier-wide correlation effects from individual processing type correlations by calculating partial correlation coefficients. This extraction process isolates genuine direct call relationships from spurious correlations caused by common tier-wide trends, making it easier to identify true call relationships.
Data Source
AI summary
A performance diagnostic server acquires time series data that indicates the number of executions per time from a multi-tier system and calculates, on the basis of the time series data, a correlation coefficient of processing types that are executed by a server in each tier. The performance diagnostic server calculates, from a correlation coefficient of processing types, a partial correlation coefficient by removing the effect of the trend of the entire multi tiers and updates the correlation coefficient of the processing types on the basis of the calculated partial correlation coefficient.