10 Years of Distributed Postgres® High Availability

From hand-drawn sketches to always-on infrastructure

Ten years ago, two things about enterprise infrastructure were becoming clear. Businesses would come to demand more from their databases than any single machine could deliver, and downtime would no longer be an accepted operating cost.

Today, both are nonnegotiable and AI is raising the stakes: Agents are becoming a primary consumer of enterprise data, querying continuously and at machine speed across every region a business operates in. Meeting that demand takes distributed high availability—running your database in multiple locations at once, so when one fails, the others keep serving traffic without interruption.

That’s what EDB Postgres AI (EDB PG AI) is built for. It turns standard PostgreSQL into a distributed, always-on platform you can run on premises, in hybrid environments, or across any cloud. It wasn’t always a platform, though—it started as an open source replication project, and this August marked 10 years since that project’s first production release. This is the story of how it developed, and a celebration of the people who built it.

Continuous availability and massive scale

Before the history, it helps to see what this looks like in practice. For systems such as global payment processing, emergency services, and utility grids, a second of delay is effectively an outage. Let’s take payment processing as an example.

Payment systems giant ACI Worldwide processes more than $14 trillion in payments and securities transactions every day for 19 of the world’s top 20 banks. That volume would challenge any company, but ACI has to handle it under strict regulatory and performance requirements with no downtime. Behind every one of those transactions is a person: someone buying groceries, making payroll, or closing on a house. A database that pauses is a person who is stuck. Jeremy Wilmot, ACI’s chief product officer, noted that “in the payments world, as you can imagine, you’ve got to be 24/7/365, particularly with transaction volume reaching billions per week.” Soon, he added, that will rise to billions per day, and “we’ll need to be able to deal with it.” Since adopting EDB PG AI, ACI hasn’t had to worry about database downtime.

The origin story: Bi-Directional Replication

Building the framework to provide always-on availability for mission-critical workloads was no simple task. It posed complex engineering challenges that required a village of brilliant people to dive into physics, distributed computing theory, and fault injection techniques. Credit for laying the foundation goes to the former team at 2ndQuadrant and the user community that built Bi-Directional Replication (BDR)—the first open source active-active replication system for PostgreSQL to reach full production status. Many members of the original team that built it are still building at EDB, and we had a chance to speak with them about the origin story.

Two of BDR’s original creators are still building at EDB. Gianni Ciolli (center), VP of Practice Solutions; and Petr Jelinek (right), VP, chief architect of Postgres Distributed, shared insights with Danny Navo (left), EDB’s database product marketing manager, on BDR’s origins and the importance of collective feedback in the Postgres community.

BDR 1.0 was released on August 11, 2016. However, it wasn’t actually the first release. According to Gianni Ciolli, VP of Practice Solutions at EDB, “This is the tenth anniversary of version 1.0, but our solution was already in production for two years.” Like many great products, it all began with customer feedback. In this case, a financial services customer wanted a distributed database based on PostgreSQL. Simon Riggs, the late EDB Postgres Fellow, began the initial design and project planning in late 2011. Andres Freund joined in and eventually became the main developer. Petr Jelinek, VP, chief architect of Postgres Distributed at EDB, recalls that the team had a working prototype in 2012.

When we asked Craig Ringer, senior staff SDE at EDB, about the original vision, his message was clear: “Distributed asynchronous active-active for Postgres. Not a seamless database you could pretend was a single node. Something [that] apps would have to understand but that could offer really impressive distributed active-active replication.” He is most proud of helping make distributed Postgres an operational reality for users, recalling “polishing rough edges and making it possible to operate without deep code-level knowledge of every part.” Craig shared original hand-drawn sketches.

Without BDR, there is no distributed Postgres at EDB. What that team solved—active-active replication, conflict resolution, surviving the loss of an entire region—became the groundwork EDB has built on ever since.

A decade strong

Today, EDB PG AI represents a decade of battle-tested database technology that continues to rapidly evolve and mature. The largest enterprises across banking, finance, telecom, and national infrastructure are shifting their mission-critical applications to EDB PG AI. The platform delivers massive scale and reliability, supporting over 80 financial institutions. The product has carried a few names since 2016, but the mission has remained the same. Always on. Everywhere.

EDB is committed to giving enterprises one platform instead of a patchwork of databases, add-ons, and third-party tools. We build new capabilities into the database rather than bolting them on. In 2025, we introduced Connection Manager, which routes traffic automatically so applications run smoothly during spikes and maintenance windows without customers ever noticing. In 2026, we added Quorum Commit, ensuring transactions like customer orders or payments are never lost or contradicted if a data center goes offline. We ship features like these with the database, so there is nothing extra to install, license, or staff for. Fewer vendors, fewer moving parts, one comprehensive solution for global, always-on applications.

The product has come a long way since the early days when one of the first production installs took two developers a full week to complete. A single CLI command handles it now. At its core, EDB PG AI is powered by the latest generation of what was launched 10 years ago. However, what began as a fix for multi-region availability is now the bedrock of the modern enterprise and the agents querying it around the clock.

So let’s raise a glass to 10 years of distributed high availability on PostgreSQL, and to the innovation and teamwork that got us here from hand-drawn sketches. See you in 2036.

If you want to learn more about achieving continuous availability with EDB PG AI, contact us to speak with EDB experts or visit our website.

If you want to dive deeper into our distributed database solution, review our documentation and the latest release notes.