How We Pushed CDC into Postgres

The Story At A Glance

• • Snowflake developed a snowflake_cdc extension to push data changes directly from PostgreSQL to object storage.


• • This architecture moves change capture inside the database process rather than using external pull-based services.


• • The system uses Apache Iceberg logs and a finite state machine to ensure deterministic data merging.

Context
Traditional Change Data Capture relies on external connectors like Debezium to pull data from Write-Ahead Logs. Snowflake is inverting this model to reduce operational complexity and latency by embedding the logic within the source database.

Christian Perspective