Vector Database Performance Challenges, A CaseStudy
Presented by:
Robert Bernier
Robert's experience extends several decades. His first experience was playing hangman on a DECwriter shortly after man first landed on the moon. His foray into commercial applications was programming Fortran, via punchcards, on an IBM 360 which in those days had 4MB RAM. Over the ensuing years he progressed into the first generation of workstations among which included the IBM 8080 XT, and AT. Robert discovered Linux and the opensource world in the 90s. By the early 2000s his primary focus had become all things PostgreSQL and has since then worked in various sized companies both small and large including Fortune 50 companies.
Robert is the Senior PostgreSQL Consultant at Percona. In his spare time he enjoys all things nature including hiking, camping, canoing and trying not to fall off the side of mountains here in the Pacific Northwest.
No video of the event yet, sorry!
AI brings vector data types to our attention like lightening warns us of the impending thunderclap in a storm. And like thunder, we don't know what it means until its too late. Despite the rapid advances developing pg_vector, an improperly designed system is more than likely to occur resulting in slow performance.
This case-study is based upon the challenges developing a solution manipulating media based recommendations vectors in a high performant fashion.
- Date:
- Duration:
- 20 min
- Room:
- Conference:
- Postgres Conference: 2026
- Language:
- Track:
- Essentials
- Difficulty:
- Medium