Microsoft patents database system that self-corrects query estimates
Microsoft has patented a method for database systems to improve how they predict the amount of data a query will return. The system compares its initial guess against the actual result, and if the guess is consistently wrong, it automatically switches to a more accurate prediction model.
In the filing’s own words
Methods, apparatuses, and products for cardinality estimation feedback for expressions in a database query, including: calculating, for one or more expressions of a database query, a default estimated cardinality and an actual cardinality, wherein the default estimated cardinality is based on an assigned cardinality estimation (CE) model comprising a default CE model, wherein the actual cardinality is based on an execution of the database query; calculating, for the one or more expressions of the database query and based on the default estimated cardinality and the actual cardinality, a default cardinality error; selecting, for a particular expression of the one or more expressions of the database query, based on the default cardinality error for the particular expression exceeding a first threshold, a recommended CE model; and assigning the recommended CE model as the assigned CE model for the particular expression based on an improvement in one or more respective recommended cardinality errors for the particular expression relative one or more respective default cardinality errors for the particular expression exceeding a second threshold, wherein the one or more respective recommended cardinality errors for the particular expression are based on one or more recommended estimated cardinalities calculated using the recommended CE model.
What exactly they claim (claim 1)
1. A method of cardinality estimation feedback for expressions in a database query, comprising: calculating, for one or more expressions of a database query, a default cardinality error associated with a default cardinality estimation (CE) model based on an estimated cardinality produced by the default CE model and an actual cardinality from execution of the one or more expressions; selecting, for a particular expression of the one or more expressions of the database query, based on the default cardinality error for the particular expression exceeding a first threshold, a recommended CE model as a potential replacement for the default CE model for the particular expression; monitoring cardinality errors of the default CE model and cardinality errors of the recommended CE model based on further executions of the particular expression; and assigning the recommended CE model to the particular expression based on reduction in the cardinality errors of the recommended CE model with respect to the cardinality errors of the default CE model exceeding a second threshold.
Facts
- Company
- Microsoft (Microsoft Technology Licensing, LLC)
- Official title
- Cardinality estimation feedback for expressions in a database query
- Patent number
- US12762122B2
- Application number
- 19083395
- Filed
- Mar 18, 2025
- Granted
- Oct 13, 2026
- Inventors
- 5
- Claims
- 20
Classified under
G06F 16/24545 Physics & computingMore from Microsoft in AI & software
- Microsoft patents method to check AI answers against verified facts Oct 20, 2026
- Microsoft patents AI that presents PowerPoint slides for you Sep 3, 2026
- Microsoft patents method to check if LLM answers are based on real data Sep 10, 2026
- Microsoft patents system to blend multiple AI agents into one chat Oct 1, 2026
- Microsoft patents system that answers questions using your screen history Oct 15, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.