AI/ML-Driven Query Optimization Using Balsa and LEON

This talk first explores PostgreSQL's query optimization capabilities, such as statistics updates, query and index optimization, and database configuration tuning. It then delves into machine learning-based, automated query optimization by introducing BALSA and LEON. LEON is based on BALSA, and both are innovative frameworks from academia. LEON combines machine learning with traditional rule-based techniques. BALSA hooks into the PostgreSQL query planner with the pg_hint_plan extension. LEON involves plan ranking and cost model calibration, which needs to change the PostgreSQL source code, which in turn requires recompiling it. The talk will conclude with a Q&A session, providing a platform for discussion on the present and future of automated query optimization. After the talk, there will be an opportunity for a free exchange of ideas about the world's most advanced open source multi-model database.

We would like to thank VSHN for hosting us at their offices in Zurich. After the presentation, interested participants can continue the discussion over a drink and snacks at the nearby bar/restaurant 'Amboss Rampe'.

If you would like to attend, please register by emailing [events@swisspug.org](mailto:events@swisspug.org).

Speaker: Stefan Keller
Location: VSHNtower, Neugasse 6, 8005 Zürich
Language: German (Slides in English)
Fees: This event is free of charge and open to the public.