Request Load Model Generation via Estimation Distribution Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating load information in computer systems are inefficient, as they rely on brute-force comparison of classification methods and exhaustive regression analysis, making it difficult to generate a valid load model efficiently.
Innovation Solution
A load estimation system that includes a management server connected to an object system, which generates a request load model by correlating load information with classified request information and selects classifications for sub-classification based on estimation distribution information, allowing for efficient division and prediction of load distributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If brute-force comparison of classification methods is employed to evaluate validity, then comprehensive evaluation is achieved, but evaluation efficiency deteriorates
Solution Approach 1:
The patent applies preliminary action by calculating estimation distribution information (such as variance or standard deviation) for each request classification before performing brute-force comparison. This preliminary calculation identifies classifications with high load variability that are most likely to impact overall system load, allowing the evaluation process to focus on these critical classifications first. The method computes estimation distribution information in advance, then uses it to prioritize which classifications require detailed evaluation, thereby maintaining comprehensive validity assessment while significantly improving evaluation efficiency.
2Manufacturing precision
If exhaustive regression analysis is performed for each processing unit of transaction, then complete model construction is achieved, but construction efficiency deteriorates
Solution Approach 1:
The patent applies preliminary action by performing regression analysis on aggregated request classifications first, before proceeding to detailed analysis of individual processing units. The method initially constructs a model using grouped classification data to identify significant patterns, then selectively performs exhaustive regression analysis only on classifications that show substantial impact on system load. This two-stage approach ensures complete model construction accuracy while avoiding unnecessary computational expenditure on minor classifications.
Solution Approach 2:
The patent applies segmentation by dividing the model construction process into multiple stages: first aggregating requests by classification and performing regression analysis at this higher level, then segmenting down to individual processing units only for classifications that require detailed analysis. This hierarchical segmentation allows the system to maintain comprehensive model accuracy while reducing overall computational complexity by avoiding exhaustive analysis at the finest granularity level for all classifications simultaneously.
3Reliability
If classification methods with small influence on load are included in comparison, then comprehensive evaluation is achieved, but evaluation time increases
Solution Approach 1:
The patent applies preliminary action by calculating estimation distribution information (such as variance, standard deviation, or coefficient of variation) for each request classification before the brute-force comparison process. Classifications with small influence on load exhibit low estimation distribution values, indicating minimal variability and impact on system performance. The method uses these pre-calculated values to identify and prioritize classifications for detailed evaluation, maintaining comprehensive evaluation completeness while significantly reducing evaluation time by focusing resources on high-impact classifications.
Data Source
AI summary
To efficiently construct a request load model that enables estimation of load information for a system on the basis of request information. A load estimation system comprising a management server that is connected to at least one object system, whereinthe management server comprises model generation means that generates a request load model in which load information for the object system is correlated with a classification of request into which request information for the object system is classified, andthe model generation means selects, in a process of the classification of the request information, the classification of request to be an object for sub-classification, based on estimation distribution information about the load information for each of the classification of request.