Guide

One-way database replication for OT networks

Updated November 17, 2026 · 7 minute read

Plants and utilities keep valuable data in databases: production records, alarms, meter readings, work orders. Business teams want that data for reporting and analysis, but connecting their systems straight to the production database creates exactly the two-way path a protected network is trying to avoid. One-way replication solves this: a copy of the data flows out, and nothing flows back.

The idea

A sender on the protected (high) side reads changes from the source database. It transmits them across a one-way link. A receiver on the outside (low) side applies the changes to a replica database. Reports, dashboards and analysts query the replica, never the original.

Why this is harder than it sounds

Normal database replication relies on conversation. The replica tells the source what it has received, asks for what is missing, and reports errors. Across a data diode none of that is possible, so the design has to answer some hard questions.

  • Ordering. Changes must be applied in the order they happened, or the replica ends up in a state that never existed.
  • Exactly once. A change applied twice, or not at all, silently corrupts the copy. The receiver must recognize duplicates and detect gaps.
  • Loss. Packets on a one-way link can be lost. Forward error correction repairs most losses by itself, but the design must also handle what repair cannot fix.
  • Recovery. If the receiver knows it has missed something, it cannot ask for it again. A well-designed system needs a way to get back to a known-good state without a request.

How change data capture fits

Instead of copying whole tables over and over, change data capture (CDC) reads the database's own record of what changed. For example, SQL Server's CDC feature records inserts, updates and deletes. The sender turns those changes into an ordered stream. This is efficient, it keeps load on the production database low, and it gives each change a position that the receiver can use to check order and spot gaps.

What a safe receiver does

  • Checks every batch end to end before applying any of it.
  • Applies changes in order, exactly once, and recognizes replays.
  • Stops and raises an alarm when something is missing, instead of silently drifting out of date.
  • Recovers by taking a fresh snapshot that the sender broadcasts, so the replica returns to a correct state without ever asking for it.
  • Keeps a verifiable record of what it applied.

Practical tips

  • Start with the tables you really need. Replicating less data is faster and easier to verify.
  • Treat the replica as read-only. Give the reporting tools read access only, and keep it on the less trusted network.
  • Monitor lag and alarms. A replica that is quietly hours behind is worse than one that says it is behind.
  • Test failure on purpose. Unplug the link, restart the receiver, drop packets, and watch what the system does before you rely on it.

What DiodeGuard does

DiodeGuard reads from SQL Server using change data capture, any database with an ODBC driver, or SQLite on the high side. It replicates into SQL Server, PostgreSQL or SQLite on the low side, ordered and exactly once, with automatic snapshot recovery and an alarm when anything is missing. See how it works and the Security page for exactly what is and is not guaranteed.

New to the topic? Start with what is a data diode, or compare approaches in data diode versus firewall.

Try one-way replication on your data

Start a free 14-day trial of DiodeGuard. No card needed.

Start free trial