Scale out, not up.
All on Postgres®.
When writes outgrow a single server, horizontal scaling spreads data and writes across groups of servers, each with its own automatic failover, so EDB Postgres AI Distributed (PG AI Distributed) grows with every customer and every agent. Now in preview.
Your database was sized for people, not agents. Agents never log off.
AI agents work every hour of every day, and every task is another write. Read replicas add copies of your data, not write capacity, so a single server eventually fills up. Until now, the options were a bigger server or a move off Postgres.
What you get with horizontal scaling
See EDB Postgres AI scale in action.
Watch database scaling in three acts. First it scales vertically, then data groups become shard groups, and finally, transactions are spread across groups of servers with horizontal scaling.
Break through the write ceiling with shard groups.
Pick a key you already have
Mark a table as distributed using standard Postgres hash partitioning. Rows are split across groups by customer, tenant, or device ID.
Writes go straight to the correct group
Each session connects to the group that holds its data, and the transaction commits there. The other groups aren't involved.
Nodes can fail safely
A node can fail. The rest of your system doesn't notice. The affected group elects a new leader in seconds, and there is no central coordinator to take everything down.
Capacity grows with your workload
Adding a group raises the write ceiling. At general availability, groups can be added to a running cluster while data rebalances online.
Proof in production
years of distributed Postgres expertise
data loss
financial institutions run always-on
uptime with distributed HA
Customers using PG AI Distributed today
"EDB Postgres AI provided what we needed in a single lower-cost solution, without any additional add-ons."
Processes payments for 19 of the world’s top 20 banks: Systems run 24/7/365 and downtime isn’t an option. EDB PG AI delivers always-on uptime across ACI’s data centers.
"Performance has remained consistent, reliability has improved, and we now have a database that can support all our future growth plans."
Modernized its fleet-tracking platform without taking the database offline. Zero downtime through upgrades and growth, on a foundation built for what’s next.
Ready to scale past your write ceiling?
The most important decision in a horizontally scaled database is the key you split your data on, and it's one you won't change later. Our engineers can help you choose it against a real workload before general availability.