The Building Blocks for Self-Driving PostgreSQL | Citus Con: An Event for Postgres 2022 Keynote

Video of a keynote talk presented by Andy Pavlo of CMUDB at Citus Con: An Event for Postgres. Abstract: Over the last 50 years, people have built advisory tools to assist in all aspects of database tuning and optimization. Most of this work is incomplete because it requires humans to make decisions about changes to the database—plus the tools are reactive, fixing problems only after they occur. A “self-driving” database aims to overcome these limitations by anticipating future workload trends, automatically deploying optimizations without human intervention, and learning from its environment to handle new situations. And a self-driving database can be optimized in ways that are not possible today, because the complexity of databases has surpassed the abilities of human experts.

In this talk during the Americas livestream at Citus Con, Andy Pavlo covers state-of-the-art in autonomous DBMSs—namely what is possible today and what problems remain—as well as the efforts at Carnegie Mellon to build a true self-driving DBMS in the NoisePage and OtterTune projects. We’ll then explore how PostgreSQL already has most of the instrumentation and extension hooks needed to control its runtime behavior via ML-based planning components. We’ll finish with Andy’s “wish-list” of features for PostgreSQL—the building blocks—that can help to bring self-driving PostgreSQL closer to reality.

Andy Pavlo is an Associate Professor of Databaseology in the Computer Science Department at Carnegie Mellon University. His (unnatural) infatuation with database systems has inadvertently caused him to incur several distinctions, such as VLDB Early Career Award (2021), NSF CAREER (2019), Sloan Fellowship (2018), and the ACM SIGMOD Jim Gray Best Dissertation Award (2014). He is also the CEO & co-founder of the OtterTune database tuning start-up (2020).

► Video bookmarks:
⏩ 00:00 Introduction
⏩ 02:05 What is a self-driving database?
⏩ 08:32 Does Postgres have what we need for a self-driving database?
⏩ 10:31 Workload capture
⏩ 16:34 System telemetry
⏩ 20:38 What-if API
⏩ 25:28 Action APIs
⏩ 29:01 So, does Postgres have what we need?

✅ Learn more:
Watch more Citus Con talks: https://aka.ms/cituscon-playlist

📕 Everything you need to know about Citus Con: An Event for Postgres can be found at: https://aka.ms/cituscon

📌 Let’s connect:
Twitter – @CitusCon, https://twitter.com/CitusCon

#CitusCon #PostgreSQL #Database