The best database talks don’t start with “here’s how it’s supposed to work.” They start with “here’s what broke.”
Percona Live is coming to the Bay Area on May 11-13, 2027 at the Computer Science Museum, and the call for papers is open until March 5th 2027. If you run, build, migrate, or secure open source databases, we want your story on stage, especially the parts that went sideways.
Open source databases are built in the open. The knowledge of running them in production should be shared the same way. That’s what Percona Live is for.
This is a practitioner conference. There are no vendor pitches and no product tours, just deep talks on real systems: the query that took down checkout, the migration plan that survived contact with production, the config change that cut your cloud bill in half. First-time speakers and seasoned veterans get the same stage. If you’ve solved a hard problem, you have a talk.
What we’re looking for
We welcome talks across MySQL, PostgreSQL, MongoDB, MariaDB, Valkey, and the wider open source database ecosystem. This year, we especially want submissions on these themes:
Cost optimization: Right-sizing, storage tiering, consolidation, and moving off proprietary licenses or managed services. Show us what the savings looked like six months later, not just on the spreadsheet.
Migration: Oracle to PostgreSQL, major version upgrades, on-premises to cloud (and back again), and cross-engine moves. Tell us what your runbook missed.
Sovereignty and compliance: Data residency, privacy regulation, and control over where your data lives and who can reach it. Show us how open source kept you out of lock-in when the rules changed.
Databases for AI agents: Agents are fast becoming a primary consumer of your data. How are you preparing? Think schema and metadata design for machine readers, semantic layers, vector search, retrieval, query guardrails, and isolating agent workloads so they don’t flatten your primary.
Identity and security for agentic access: When an agent queries your database, who is it? We want talks on authenticating non-human identities, scoping permissions, auditing agent activity, and stopping prompt injection before it reaches your data.
Session formats
30 minutes, including Q&A: One focused problem, one solution, and the results. Good for a single incident, benchmark, or technique.
60 minutes, including Q&A: Room for depth. Good for end-to-end migrations, architecture walkthroughs, or hands-on demos.
Leave about 10 minutes for questions in either slot. Tell us which length you want when you submit. If your topic fits the other format better, we may suggest it.
Think your idea is too small?
Submit it anyway. A narrow topic with real detail beats a broad overview every time, and a 30-minute slot is built for one. One production incident, one benchmark, or one migration done properly is a strong talk.
Submissions close March 5th 2027.
If you haven't logged in before, you'll be able to register.
Using social networks to login is faster and simpler, but if you prefer username/password account - use Classic Login.