Inferring Configuration Effects on Computing Machine Performance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying configuration settings that affect computing machine performance under stress rely on trial-and-error approaches, lacking a systematic and data-driven method to determine optimal settings.
Innovation Solution
Collecting configuration and performance data from computing machines, identifying periods of stress, grouping machines by stress profiles, and inferring the effects of configuration points on performance to determine a baseline set of settings that positively impact performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If trial-and-error approach is used to identify configuration settings, then configuration settings can be identified, but the process is inefficient and lacks systematic optimization
Solution Approach 1:
The system collects performance data from computing machines, analyzes it to identify stress periods, and uses this feedback to infer the effects of configuration points on performance. This closed-loop feedback mechanism replaces trial-and-error with data-driven optimization, systematically identifying configuration settings that improve performance during stress periods.
Solution Approach 2:
The system enables computing machines to self-optimize their configuration settings by automatically collecting their own performance data, analyzing it locally or remotely, and inferring optimal configuration points without requiring external trial-and-error intervention. This self-service approach accelerates configuration optimization.
2Reliability
If configuration settings are changed to improve performance, then performance can be enhanced, but negative impacts may occur if settings are not optimized
Solution Approach 1:
The system systematically varies and analyzes configuration parameters (such as registry keys, file versions, number of network cards) to determine their effects on performance during stress periods. By measuring performance changes in response to parameter changes, the system identifies settings that enhance reliability without causing negative impacts.
Solution Approach 2:
The patent replaces manual trial-and-error configuration tuning with an automated data analysis system that uses performance monitoring and statistical inference to determine optimal settings. This substitution eliminates guesswork and reduces the risk of introducing harmful configuration changes.
3Productivity
If data collection and analysis is performed to infer configuration effects, then systematic optimization is achieved, but system complexity increases
Solution Approach 1:
The system performs multiple functions using a unified data collection and analysis framework: it monitors performance metrics, identifies stress periods, collects configuration data, and infers configuration effects all within the same system architecture. This multi-functionality reduces overall system complexity compared to separate specialized systems for each function.
Data Source
AI summary
Configuration data and performance data can be collected from computing machines running a target program. Periods of stress for the computing machines can be identified using the performance data, and a set of the computing machines can be grouped under a stress profile using the performance data. One or more configuration points can be identified on the set of machines, and an effect of each of the configuration point(s) on performance of the set of machines can be inferred using the configuration data and the performance data. The inferred effect(s) of the configuration point(s) can be used to determine a baseline set of configuration settings.