Synthetic Data Generation for Performance Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data generation methods are inadequate for performance evaluation of applications, as they primarily focus on testing correctness rather than performance evaluation, and often lack the ability to generate large volumes of synthetic data that accurately mimic real data characteristics.
Innovation Solution
A computer-implemented method that identifies a source dataset, extracts its schema and value models, determines data dependencies, and generates synthetic data based on these analyses to create a data specification file, allowing for the generation of synthetic data that matches the performance characteristics of the source data, including varying sizes and complexities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing data generation methods are used, then data can be generated for testing correctness, but the data is not useful for performance evaluation
Solution Approach 1:
The system performs preliminary analysis of the source data to extract value models, data dependencies, and schema information before generating synthetic data. This preliminary action ensures that the generated data inherently possesses the performance characteristics of the source data, making it suitable for performance evaluation from the outset rather than requiring post-generation adjustments.
Solution Approach 2:
The system creates synthetic data that copies the essential performance-relevant characteristics of the source data through value models and dependency relationships. Instead of copying the actual data values, it copies the structural and statistical properties that determine performance behavior, enabling realistic performance evaluation without using real data.
2Measurement precision
If synthetic data is generated without analyzing data dependencies, then generation is simpler and faster, but the synthetic data does not accurately mimic real data characteristics
Solution Approach 1:
The system segments the complex task of synthetic data generation into distinct analytical components: extracting value models for individual columns, identifying data dependencies between columns and tables, and analyzing schema constraints. This segmentation allows each aspect to be analyzed and modeled separately, improving accuracy while making the overall process more manageable and systematic.
Solution Approach 2:
The system performs comprehensive preliminary analysis of data dependencies, value distributions, and schema constraints before the actual data generation process. By capturing these characteristics in advance through automated analysis, the system ensures high accuracy in the generated data without requiring complex real-time adjustments during generation.
3Quantity of substance
If large volumes of synthetic data are generated for performance evaluation, then performance testing is more realistic, but existing methods cannot generate sufficient data volume with proper characteristics
Solution Approach 1:
The system creates a universal data generation framework based on extracted value models and dependency relationships that can generate synthetic data in any required volume while maintaining consistent characteristics. The same value models and dependency rules that capture the essence of the source data can be applied repeatedly to generate arbitrarily large datasets, making the system both versatile in volume and adaptive in characteristics.
Data Source
AI summary
The present disclosure describes methods, systems, and computer program products for generating data for performance evaluation. One computer-implemented method includes identifying a source dataset from a source database, extracting a schema defining the source database, analyzing data within the source dataset to generate a value model, the value model describing features of data in each column of each table of the source dataset, analyzing data within the source database to determine data dependency, and generating a data specification file combining the extracted schema, the value model, and the data dependencies.