Data Pipeline Output Validation Using Data and Schema Fingerprints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cross-validation of data processing systems, such as data analysis systems, is challenging due to hardware and software incompatibilities, leading to data mismatches, corruption, loss, or inaccuracy during migration, and manual validation is time-consuming and resource-intensive, especially for large datasets.
Innovation Solution
A method involving data and schema comparison through aggregation of data tables to generate fingerprints, followed by a summary calculation to assess similarity between the systems, allowing for automated validation without line-by-line comparison.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct comparison of source output table and migrated output table is performed, then validation accuracy is improved, but transmission requirements and resource consumption increase significantly
Solution Approach 1:
The patent extracts only the essential validation information from the complete output tables by computing aggregation values (such as sums, counts, or other summary statistics) for each column. This allows validation to be performed on a much smaller extracted dataset rather than transmitting and comparing entire large tables, thus reducing data transmission volume while maintaining validation capability.
Solution Approach 2:
Instead of performing complete line-by-line comparison of all data rows, the patent applies partial action by comparing only aggregated values derived from the data. This partial comparison approach is sufficient to detect significant migration issues while avoiding the resource-intensive full comparison, effectively balancing validation thoroughness with resource consumption.
2Measurement precision
If manual checking and validation is performed, then validation thoroughness is improved, but time and resource consumption increase
Solution Approach 1:
The patent implements self-service validation by automatically computing aggregation values from both source and migrated output tables and performing comparative analysis without human intervention. The system generates validation reports that automatically identify mismatches, allowing thorough validation to be performed quickly and efficiently without requiring manual checking of individual data rows.
Solution Approach 2:
The patent replaces manual mechanical validation processes with automated computational systems. Instead of human operators manually comparing data tables, the system uses processors to automatically compute aggregation values, compare them, and generate validation results, thereby maintaining thoroughness while dramatically reducing validation time and human resource consumption.
3Productivity
If data tables are aggregated into smaller tables, then data transmission and comparison efficiency are improved, but validation detail level may be reduced
Solution Approach 1:
The patent applies local quality by computing aggregation values for each column individually rather than treating the entire table as a single unit. This allows the validation to maintain sensitivity to column-specific issues while benefiting from the efficiency of aggregated comparison. Each column's aggregation value serves as a localized summary that preserves important validation information for that specific data dimension.
Data Source
AI summary
A method is provided for output validation of data processing systems, performed by one or more processors. The method comprises performing a data comparison between a first data table and a second data table to determine a data differentiating table, wherein the first data table is based on an output of a first data pipeline, and wherein the second data table is based on an output of a second data pipeline; performing a schema comparison between the first data table and the second data table to determine a schema differentiating table; generating a first output validation score based on the data differentiating table; generating a second output validation score based on the schema differentiating table; and generating a summary comprising both the first and second output validation scores.


