Materialized Views: Performance Shortcut or Operational Trap?
Presented by:
Yu Lung Law
Yu Lung Law is a software engineer with Bloomberg's Data Platform Engineering team. He is responsible for maintaining and scaling the team’s configuration-driven data ingestion pipeline. He originally joined the company's Data department in 2012, where he worked for a decade as a data analyst & data engineer prior to transferring into the firm's Engineering department. Yu Lung earned his master's degree in computer science from the University of Pennsylvania.
Christopher Hong
Christopher Hong is a Software Engineering Team Lead in Bloomberg's Data Platform Engineering group. His team’s goal is to build reusable infrastructure to enable the company's data analysts around the globe to easily bring valuable datasets into Bloomberg and to make them available to clients via API. Throughout his 12-year career at Bloomberg around data, he has built various ETL systems using a variety of open-source technologies to successfully make hundreds of datasets available to clients.
Outside of Bloomberg, Christopher regularly mentors first-generation college students through Project Basta, a non-profit organization that aims to bridge the employment gap and build careers for first-generation students. As a first-generation college graduate himself from Brooklyn, New York, Christopher earned his master's and bachelor's degrees in electrical engineering from Cooper Union, where he is also an adjunct professor teaching computer science, software engineering, and large-scale systems to future generation of engineers.
No video of the event yet, sorry!
Materialized views are often the go-to solution for speeding up complex queries in PostgreSQL, especially for heavy joins and time-series aggregations. But while they can deliver impressive performance gains, they also introduce hidden operational costs—long refresh times, slower recovery, and increased fragility as data grows.
In this talk, we’ll walk through several real-world examples where materialized views were the right choice, how they later became a bottleneck as data and usage evolved, and how we handle these trade-offs in our systems. We’ll also discuss safer usage patterns for materialized views and compare alternative approaches.
Attendees will leave with a practical framework for deciding when to use materialized views—and when to reach for something else—before today’s performance optimization becomes tomorrow’s operational problem.
- Date:
- Duration:
- 20 min
- Room:
- Conference:
- Postgres Conference: 2026
- Language:
- Track:
- Ops
- Difficulty:
- Easy