- Explanation of the problem of inaccurate query predictions due to correlated predicates
- Overview of traditional approaches, including pre-computed summary structures and opportunistic methods
- Introduction of CORDON as a new approach that leverages workload data to learn about query predictability
Section 2: Algorithm Description
- Detailed explanation of the CORDON algorithm, including the steps involved in detecting applicable constraints, generating augmented training samples, and end-to-end training
- Discussion of how the algorithm uses logic loss functions to enforce domain knowledge and ensure accurate predictions
Section 3: Training and Serving
- Explanation of how the trained model is used to serve queries from the database management system (DBMS)
- Discussion of how CORDON can be used to improve query performance in real-world scenarios
Conclusion
In this article, we presented CORDON, a new approach to improving the performance of database systems by leveraging workload data to learn about query predictability. By using workload results to inform the learning process, CORDON can provide more accurate predictions and better performance than traditional methods. We believe that CORDON has the potential to make a significant impact in the field of database management, and we look forward to further researching and developing this technology in the future.
