Web System Parameter Optimization via Adaptive Batch Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for optimizing web system configuration parameters are time-consuming and subjective, often requiring extensive measurement of all parameter sets, which can lead to inaccurate performance evaluation and inefficient optimization.
Innovation Solution
A method that selectively measures performance for a subset of batch jobs, calculates an evaluation value based on the difference between measurement values for a test parameter set and an optimal parameter set, and adjusts the evaluation range to determine performance improvement or deterioration, allowing for efficient replacement of optimal parameter sets and repeated measurements until performance deterioration is detected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all parameter sets are measured to ensure accurate optimization, then measurement precision is improved, but loss of time increases enormously
Solution Approach 1:
The patent applies partial action by measuring performance for only a subset of parameter sets rather than all possible combinations. It uses intelligent sampling and evaluation to identify promising parameter sets without exhaustive measurement, thereby reducing optimization time while maintaining acceptable accuracy.
Solution Approach 2:
The system performs self-service through automated evaluation and selection of parameter sets. It autonomously determines which parameter sets to measure based on preliminary analysis and performance criteria, eliminating the need for manual exhaustive testing while ensuring optimal configuration identification.
2Ease of operation
If a fixed period of time is used for performance measurement, then ease of operation is improved, but measurement precision deteriorates when the time is too short or too long
Solution Approach 1:
The patent applies dynamics by making the measurement period adaptive rather than fixed. The measurement time is dynamically adjusted based on the specific parameter set being evaluated, system response characteristics, and performance convergence criteria, ensuring both operational simplicity and measurement precision.
3Device complexity
If parameter sets are selected based on designer intuition, then device complexity is reduced, but reliability deteriorates due to subjectivity
Solution Approach 1:
The patent implements feedback mechanisms where performance measurement results are systematically collected and used to guide subsequent parameter set selections. This objective feedback loop replaces subjective designer intuition with data-driven decision-making, ensuring reliability while maintaining manageable process complexity.
Solution Approach 2:
The system systematically varies and evaluates multiple configuration parameters to identify optimal settings. By automating parameter exploration and evaluation, it objectively determines best configurations without relying on designer subjectivity, thereby improving reliability while keeping the process systematic and manageable.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The method includes the steps of: storing a plurality of parameter sets; selecting one of the plurality of parameter sets as a test parameter set to be evaluated; measuring performance only for one batch job out of N (N is a positive integer) batch jobs constituting full set performance measurement for the test parameter set; and calculating an evaluation value on the basis of a difference between an integral of measurement values obtained until the performance has been measured for r (r is a positive integer smaller than N) batch jobs by using the test parameter set; and an integral of mean measurement values of the performance for the r batch jobs by using an optimal parameter set which is one of the parameter sets used in the performance evaluation having been performed; determining whether or not the evaluation value has deviated from a predetermined evaluation continuing range; and terminating the evaluation of the test parameter set on condition that it is determined that the evaluation value has deviated form the evaluation continuing range toward performance deterioration. It is preferable that the predetermined evaluation continuing range be of a width from a width W where r is equal to zero, to a width W' (0 ≤ W' < W) where r is equal to N. Furthermore, it is preferable that W' be zero.