Bottleneck Detection via Request Count Frequency Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-tier information processing systems face challenges in accurately detecting bottlenecks, as current methods struggle to identify the tier causing process delays and may overlook anomalies due to limitations in resource management and process time analysis.
Innovation Solution
An operation management apparatus that calculates the ratio of retention job request counts within a specific range to the total sum across multiple information processing apparatuses, determining a concentration indicator to identify saturated servers and detect bottlenecks based on frequency distributions and normal distribution analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional resource usage status and process time analysis methods are applied, then system operation management is simplified, but bottleneck detection accuracy deteriorates
Solution Approach 1:
The patent changes the parameter for bottleneck detection from traditional resource usage status and process time to frequency distribution characteristics of sampled counts. By analyzing whether the frequency distribution conforms to a normal distribution pattern, the system can accurately detect bottlenecks even when resource usage appears normal, resolving the contradiction between simple operation management and accurate bottleneck detection.
2Device complexity
If conventional bottleneck detection methods are used, then system monitoring is straightforward, but ability to detect anomalies in saturated systems deteriorates
Solution Approach 1:
The patent replaces conventional mechanical monitoring approaches (resource usage status, process time) with a statistical analysis approach. By substituting the monitoring mechanism to analyze frequency distribution patterns and their deviation from normal distribution, the system gains reliable anomaly detection capability while maintaining straightforward monitoring implementation.
3Measurement precision
If process time extension analysis is performed, then delay location identification is improved, but detection of false normalcy in limited-parallelism systems deteriorates
Solution Approach 1:
Instead of analyzing process time extension to identify bottlenecks (conventional approach), the patent inverts the approach by analyzing the frequency distribution of sampled counts to detect anomalies. This inversion allows the system to detect bottlenecks caused by parallelism limitations without being misled by false normalcy in process time data, preserving critical bottleneck cause information.
Data Source
AI summary
An operation management apparatus obtains a value Xi indicating the number of process requests being processed by an information processing apparatus during each sampling operation, from N samplings acquired during a specific time period from the information processing apparatus, wherein N is an integer satisfying a condition of 1≦N, and i is an integer satisfying a condition of 1≦i≦N. The apparatus determines, for a plurality of information processing apparatuses, a ratio of the sum of values Xi, each value Xi having a difference, from a maximum value of the values Xi, falling within a specific range, to the total sum of the values Xi. The apparatus detects an information processing apparatus having the ratio equal to or higher than a specific value.