Edge
PostgreSQL logo

For PostgreSQL

Postgres on Edge,
without the RDS markup

Self-hosted PostgreSQL on dedicated NVMe, with WAL archiving to S3-compatible storage and replication over private networking. Real performance, predictable bills, no managed-service guard rails.

# Provision a Postgres VM with NVMe

$ edge compute create \

--image ubuntu-24-04 --plan db-medium \

--script ./bootstrap-postgres.sh

# pgbackrest.conf — WAL → Edge Storage

[global]

repo1-type=s3

repo1-s3-endpoint=storage.edge.run

repo1-s3-bucket=postgres-wal

repo1-s3-region=eu-west

# Backup + verify

$ pgbackrest --stanza=main backup

✓ archive-push, basebackup complete

Why Postgres teams pick Edge

All the Postgres features, none of the managed-service "you can't do that" answers.

NVMe storage that doesn't throttle

Real local NVMe on every Edge VM — no per-IOPS billing, no burst credits to track. Postgres is happy when its disk is happy.

WAL archiving to S3

Pipe WAL segments to Edge Storage with `wal-g` or `pgbackrest` for point-in-time recovery — without the per-GB-month bills RDS charges for snapshots.

Replication on private networking

Run a primary + replicas on separate Edge VMs over private networking. No cross-AZ data transfer fees, no public-internet exposure.

Any version, any extension

Pin to whatever Postgres version you want. Install PostGIS, TimescaleDB, pgvector, pg_cron — anything that compiles, anything you compile yourself.

Pgbouncer / PgCat included

Run a connection pooler on the same VM (free) or a dedicated tiny one (cheap). No managed-service connection-limit panics.

No per-instance-hour markup

A db.r6g.large RDS instance is ~$220/month before storage. The same workload on an Edge VM with bigger NVMe is a small fraction of that — no licence fee.

Reference architecture

How Postgres maps to Edge

Primary on a sized-to-fit VM, optional streaming replica, WAL archived to object storage. Standard Postgres ops, modern infrastructure.

Compute

Postgres + Pgbouncer on a VM, with NVMe for the data directory

Storage

WAL archive bucket + nightly basebackups for PITR

DNS

Anycast DNS for `db.example.com` — point it at the primary

Compute (replica)

Optional second VM as a streaming replica via private networking

Indicative cost

Mid-sized production DB

~4 vCPU, 16GB RAM, 500GB storage, daily backups, 100GB WAL/day

RDS db.r6g.xlarge + storage + IOPS ~$500–800
Aurora equivalent ~$700–1,200
Edge (VM + Storage for WAL) ~$60–120

Indicative figures. Add a second VM for HA replication.

Common questions

How does this compare to RDS / Aurora?

Cheaper (no per-instance-hour, no per-GB snapshot fees, no per-IOPS bills) and faster on like-for-like hardware (real NVMe). Trade-off: you handle backups, upgrades and HA — or our Expert Services team can.

How do I do high availability?

Streaming replication to a hot standby on a second VM, plus automated failover with Patroni or repmgr. Both work cleanly on Edge VMs over private networking.

What about backups?

Use `pgbackrest` or `wal-g` to ship WAL segments and basebackups to an Edge Storage bucket. Point-in-time recovery without paying RDS' premium for the privilege.

Can I use it from other Edge resources?

Yes — connect over private networking from your app VMs (Laravel, Django, Rails, Node.js, etc.). Latency is sub-millisecond, no egress fees between Edge resources.

By Stack

Other stacks on Edge

View all stacks →

Run Postgres on your terms

30-day trial. Migrate from RDS or stand up a fresh cluster — Expert Services can plan the cutover.