What is it about?
In the era of big data, computing exact answers to analytical queries becomes prohibitively expensive. It takes seconds or even minutes to get the exact query result from a big table. This greatly increases the value of approaches that can compute efficiently approximate, but highly accurate answers to analytical queries.
Featured Image
Photo by Tobias Fischer on Unsplash
Why is it important?
DBEst, as the first model-based approximate query processing engine, is built over classical machine learning models. Specifically, regression models and density estimators are trained over the data/samples, and are used to produce approximate answers fast and efficiently. Compared with sample-based AQP approaches, lightweight models enjoys orders of magnitude savings in query response time and space overheads, while achieving better accuracy.
Perspectives
With this paper, we hope people will get familiar with approximate query processing techniques using machine learning models. By building models for specific query templates, model-based AQP methods enjoy benefits in all aspects. Machine learning also opens up a new area for approximate query processing, and there are more opportunities and challenges. We also proposed an improved AQP engine based on deep learning networks. Feel free to try them on!
qingzhi ma
University of Warwick
Read the Original
This page is a summary of: DBEst, June 2019, ACM (Association for Computing Machinery),
DOI: 10.1145/3299869.3324958.
You can read the full text:
Resources
DBEst SIGMOD Presentation
SIGMOD 2019 DBEst: revisiting approximate query processing engines with machine learning models
DBEst++ CIDR2021 Presentation
This is an improved model-based AQP engine
Demo of DBEst++
This is the demo video of DBEst++
Learned Approximate Query Processing: Make it Light, Accurate and Fast
This is an improved AQP engine, based on machine learning models.
Github Repository
This is the GitHub Repository for DBEst++.
Contributors
The following have contributed to this page
