South Bay Systems: ChatGPT Ain’t Got $%@& On Me! The Future of Automated Database Tuning

​We're excited to feature Andy Pavlo, illustrious database professor at CMU, to talk about database tuning.

​This meetup's venue, food and drinks, are generously sponsored by Databricks.

​Schedule

​5:30PM: Doors open
6:30PM: Andy's talk
7:30PM: Socializing

​Abstract

​There are many automated tools for optimizing the multitude of tunable components in databases (e.g., knobs, indexes, query plans). Since tuning a database is exceedingly difficult, nearly every existing tool specializes in optimizing a single tuning target at a time. This inability to holistically tune a database results in suboptimal performance and efficiency. Recent breakthroughs in AI, such as large-language models (LLMs), that enable semantic reasoning about a database appear to overcome this prior limitation, allowing a tuning tool to optimize all database facets simultaneously. However, it remains unclear how to utilize LLMs best for automated database tuning and whether they can surpass the performance of specialized algorithms.

​This talk presents research on a new generation of autonomous tuning agents that optimize more parts of a database in less time and at lower costs than previous methods. We will describe our automated methods for tuning nearly every possible option available in PostgreSQL: system/table/index knobs, index selection, and query plan hints. As part of this talk, we will also discuss our successes and failures in using LLMs to accelerate database tuning, as well as our road map towards achieving a true self-driving database system.

​Speaker Bio

​Andy Pavlo is an Associate Professor with Indefinite Tenure of Databaseology in the Computer Science Department at Carnegie Mellon University. He was the co-founder of the OtterTune start-up and record label, but then it died. He is currently the co-founder of "SO-YOU-DONT-HAVE-TO INCORPORATED'); DROP TABLE companies; --". His goal is to achieve a database-centric lifestyle.