Scaling PostgreSQL at OpenAI: Lessons in Reliability, Efficiency, and Innovation
At POSETTE: An Event for Postgres 2025, Bohan Zhang of OpenAI delivered a compelling talk on how OpenAI has scaled Azure Database for PostgreSQL- Flexible Server to meet the demands of one of the world’s most advanced AI platforms running at planetary scale. The Postgres team at Microsoft has partnered deeply with OpenAI for years to enhance the service to meet their performance, scale, and availability requirements, and it is great to see how OpenAI is now deploying and depending on Flexible Server as a core component of ChatGPT. Hearing firsthand about their challenges and breakthroughs is a reminder of what’s possible when innovation meets real-world needs. This blog post captures the key insights from Bohan’s POSETTE talk, paired with how Azure’s cloud platform supports innovation at scale. PostgreSQL at the Heart of OpenAI As Bohan shared during his talk, PostgreSQL is the backbone of OpenAI’s most critical systems. Because PostgreSQL plays a critical role in powering services like ChatGPT, Open AI has prioritized making it more resilient and scalable to avoid any disruptions. That’s why OpenAI has invested deeply in optimizing PostgreSQL for reliability and scale. Why Azure Database for PostgreSQL? OpenAI has long operated PostgreSQL on Azure, initially using a single primary instance without sharding. This architecture worked well—until write scalability limits emerged. Azure’s managed PostgreSQL service provides the flexibility to scale read replicas, optimize performance, and maintain high availability to provide global low latency reads without the burden of managing infrastructure. This is why we designed Azure Database for PostgreSQL to support precisely these kinds of high-scale, mission-critical workloads, and OpenAI’s use case is a powerful validation of that vision. Tackling Write Bottlenecks PostgreSQL’s MVCC (Multi-Version Concurrency Control) design presents challenges for write-heavy workloads—such as index bloat, autovacuum tuning complexity, and version churn. OpenAI addressed this by: Reducing unnecessary writes at the application level Using lazy writes and controlled backfills to smooth spikes Migrating extreme write-heavy workloads with natural sharding keys to other systems. These strategies allowed OpenAI to preserve PostgreSQL’s strengths while mitigating its limitations. Optimizing Read-Heavy Workloads With writes offloaded, OpenAI focused on scaling read-heavy workloads. Key optimizations included: Offloading read queries to replicas Avoiding long-running queries and expensive multi-way join queries Using PgBouncer for connection pooling, reducing latency from 50ms to under 5ms Categorizing requests by priority and assigning dedicated read replicas to high-priority traffic As Bohan noted, “After all the optimization we did, we are super happy with Postgres right now for our read-heavy workloads.” Schema Governance and Resilience OpenAI also implemented strict schema governance to avoid full table rewrites and production disruptions. Only lightweight schema changes are allowed, and long-running queries are monitored to prevent them from blocking migrations. To ensure resilience, we categorized requests by priority and implemented multi-level rate limiting—at the application, connection, and query digest levels. This helped prevent resource exhaustion and service degradation. Takeaway OpenAI’s journey is a masterclass in how to operate PostgreSQL at hyper-scale. By offloading writes, scaling read replicas, and enforcing strict schema governance, OpenAI demonstrated PostgreSQL on Azure meets the demands of cutting-edge AI systems. It also reinforces the value of Azure’s managed database services in enabling teams to focus on innovation rather than infrastructure. We’re proud of the work we’ve done to co-innovate with OpenAI and excited to see how other organizations can apply these lessons to their own PostgreSQL deployments. Check out the on-demand talk “Scaling Postgres to the next level at OpenAI” and many more PostgreSQL community sessions from POSETTE.Postgres speakers - POSETTE 2026 CFP is closing soon!
Guidelines for submitting a proposal to the POSETTE CFP POSETTE: An Event for Postgres is back for its 5th year, and the excitement is already building. Scheduled for June 16 – June 18, 2026, this free and virtual developer event brings together the global Postgres community for three days of learning, sharing, and deep technical storytelling. Whether you're a first-time speaker or a seasoned contributor, your story matters and the Call for Proposals (CFP) closes on February 1, 2026. If you’re considering submitting a proposal (or encouraging someone else to), in this post I will walk you through everything you need to know to craft a strong, compelling submission before the deadline arrives. 1. Key Dates to Know CFP Deadline: February 1, 2026 @ 11:59 PM PST Talk Acceptance Notifications: February 11, 2026 Event Dates: June 16 – June 18, 2026 (includes four unique livestreams, live text chat, and speaker Q&A) Schedule & sessions announced: Feb 25, 2026 Pre-record all talks: Weeks of April 20 & April 27 Tip: Add a calendar reminder, this deadline arrives quickly, and no late submissions are accepted. 2. Why Submit a Talk to POSETTE? Submitting a talk for a conference can seem like a difficult task at the start, but this guide can help you come up with potential ideas that can be used to submit a talk for the conference. Share your story with the global Postgres community Your experience, whether it’s a deep dive into query planning, a migration journey, or lessons learned from scaling can help thousands of developers. Grow your professional visibility POSETTE is a high‑reach, virtual event that enables your content to live on well after the livestream. First‑time speakers are welcomed and encouraged POSETTE is not an exclusive club. If you have a story to tell, this is a supportive, welcoming place to tell it. 3. What Makes a Strong Proposal? First‑time speaker? Don’t worry. The guidelines below cover the key elements you’ll need to craft a strong, successful proposal. Make your proposal focused, not broad: Many proposals try to cover too much. The strongest ones zoom in on a specific challenge, insight, or transformation. A narrow, well‑defined topic reads more clearly and creates a stronger takeaway for attendees. Clearly identify the target audience: State who the talk is for: Beginner Postgres developers Cloud architects DBAs focusing on performance Engineers migrating from Oracle/MySQL This helps the selection team understand fit and event balance. Demonstrate real‑world value, not generic theory: Talks rooted in hands‑on experience tend to perform best. Strong abstracts answer: What problem did we face? What did we try? What worked (or didn’t)? What can you replicate in your environment? POSETTE audiences love actionable content. 4. Show how attendees will grow from your talk: Selection committees love when speakers articulate transformation. Clarify what people will gain: “Improve query execution time by…” “Avoid common replication pitfalls…” “Design HA setups more confidently…” The reviewers want talks with practical outcomes. 5. Highlight what makes your talk unique Is your approach unconventional? Did you migrate at massive scale? Did you build or extend an OSS tool? Did you learn something the hard way? Emphasize novelty POSETTE gets many submissions, so originality matters. 6. Use a storytelling angle: Human brains love stories. Strong abstracts often follow a mini narrative: Problem Tension Turning point Solution Lessons This makes your proposal memorable and relatable. 7. Keep the abstract concise and structured: Avoid long, meandering paragraphs. A clear structure like this works well: Topic summary (one sentence) Problem + context (two–three sentences) Solution or insights (two–three sentences) What attendees will learn (one–two sentences) 4. Ideas for Topics That Work Well Not every proposal needs to be a deep internal dive real‑world stories resonate. Consider topics like: Migrating to Postgres (cloud or on‑prem) Performance tuning adventures and lessons Postgres extensions and ecosystem tooling Operational best practices, HA architecture, or incident learnings Developer productivity with Postgres Novel patterns or creative uses of Postgres internals Azure Database for PostgreSQL customer stories Community‑focused topics, such as how to start a PGDay event, how to begin contributing to open source, or how to engage with the Postgres community effectively. Look at POSETTE 2024 or 2025 talk titles to calibrate tone and depth. 5. What Happens If Your Talk Is Accepted? Good news: the speaker experience is designed to be smooth and supportive. Talks are 25 minutes long and pre‑recorded, with professional production support from the POSETTE organizing team at an agreed-upon time during the weeks of April 20 & April 27 Speakers join live text chat during the session to interact with attendees No travel required the event is fully virtual All you need is a good microphone, a quiet space, and a story worth telling. 6. How to Submit Your Proposal Here are the official links you’ll want handy: 📄 CFP Page: https://posetteconf.com/2026/cfp/ ❓ FAQ: https://posetteconf.com/2026/faq/ 📝 Submit on Sessionize: https://sessionize.com/posette2026/ Submission Checklist Before hitting "submit," make sure you have: A strong, interesting title A clear and concise abstract Defined takeaways for attendees An understanding of your target audience Submission completed before Feb 1 @ 11:59 PM PST POSETTE is built by and for the Postgres community and your experience, whether small or monumental, has the potential to help others. With the CFP deadline approaching fast on February 1, now is the perfect time to refine your idea, shape your abstract, and submit your talk. This could be the year your story gets shared with thousands. Take the leap the community will be glad you did.Calling Postgres speakers, POSETTE CFP is open until Apr 7th 2024
Call for Proposals (CFP) is open til Sun Apr 7 at 11:59pm PDT for POSETTE: An Event for Postgres, a free & virtual event. What’s your Postgres story? We’d love to see your talk proposal, whether you’re a first-time speaker, a regular on the Postgres conference circuit, or somewhere in between.
