Loading…
PrestoCon 2023 has ended
In-person
December 5-6, 2023
Learn More and Register to Attend

The Sched app allows you to build your schedule but is not a substitute for your event registration. You must be registered for PrestoCon 2023 to participate in the sessions. 

Please note: This schedule is automatically displayed in Pacific Standard Time (UTC -8). To see the schedule in your preferred timezone, please select from the drop-down menu to the right, above "Filter by Venue." The schedule is subject to change.
Wednesday, December 6 • 2:10pm - 2:40pm
Learned Query Optimization in PrestoDB - Nesime Tatbul & Dave Cohen, Intel; Ryan Marcus, University of Pennsylvania; Christoph Anneser, Technical University of Munich

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
As part of a research collaboration between Intel and Meta, we developed AutoSteer - an ML-based solution that automatically drives query optimization in any SQL database that exposes tunable optimizer knobs. AutoSteer builds on the Bandit optimizer (Bao) developed by Intel Labs and MIT DSAIL, and extends it with new capabilities to facilitate usability in disaggregated SQL systems such as PrestoDB. We successfully applied AutoSteer on PrestoDB (via both generic and custom connectors), and conducted a detailed experimental evaluation with both public benchmarks and a production workload from Meta’s PrestoDB deployments. We achieved up to 40% improvement in query performance vs. PrestoDB's native query optimizer, and contributed a new optimizer heuristic currently in use by Meta. AutoSteer comes with a visual frontend and is available open source. At PrestoCon, we would like to share our findings with the broad PrestoDB community to get feedback and promote future collaboration.

Speakers
avatar for Dave Cohen

Dave Cohen

Sr Principal Engineer, Intel
Dave has been working on large-scale, distributed data systems for over 30 years.
avatar for Nesime Tatbul

Nesime Tatbul

Sr. Research Scientist, Intel Labs and MIT
avatar for Christoph Anneser

Christoph Anneser

Mr., Technical University of Munich
Optimizing database systems adaptively at runtime.
avatar for Ryan Marcus

Ryan Marcus

Assistant Professor, University of Pennsylvania
An assistant professor at UPenn, Ryan researches applications of machine learning to data systems, as well as learned and instance optimized systems.



Wednesday December 6, 2023 2:10pm - 2:40pm PST
Boole Room
  Presto Track, Session Presentations
  • Level Any
  • Presentation Slides Attached Yes