What is it about?
Recent performance prediction models have made great advances in accurately predicting the execution time of database queries. However, the neural network models they use have higher latencies than many short-running queries themselves. See how we achieved the same accuracy with a 10,000x faster model, T3.
Featured Image
Photo by benjamin lehman on Unsplash
Why is it important?
Many modern data systems use performance estimates for important scheduling and resource scaling decisions. A slow prediction directly incurs higher query latency.
Perspectives
I built T3 to predict execution times for Umbra. While latency was outstanding from the beginning, there was a lot of work to do to get the accuracy on par with state of the art. See how decomposition into pipelines and per tuple predictions allow one to approach this problem.
Maximilian Rieger
Technische Universitat Munchen
Read the Original
This page is a summary of: T3: Accurate and Fast Performance Prediction for Relational Database Systems With Compiled Decision Trees, Proceedings of the ACM on Management of Data, June 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3725364.
You can read the full text:
Contributors
The following have contributed to this page
