The short answer
Traditionally, replicating operational transactional data from Amazon Aurora (PostgreSQL or MySQL) into Amazon Redshift required building and maintaining custom (CDC) pipelines using tools like AWS DMS, Kafka Connect, or Glue streaming jobs. These pipelines introduced operational complexity, lag, and data transformation failures.
AWS Aurora Zero-ETL integration provides a fully managed, serverless replication mechanism that moves data from Aurora directly into Redshift within seconds.
+-------------------------------------------------+
| Amazon Aurora Cluster |
| Transactional OLTP application |
+-------------------------------------------------+
|
| storage-level log streaming
v
+-------------------------------------------------+
| Zero-ETL Replication Engine |
| Serverless AWS-managed log processing |
+-------------------------------------------------+
|
v automatic continuous apply
+-------------------------------------------------+
| Amazon Redshift Cluster |
| Target read-only analytical database |
+-------------------------------------------------+How it works under the hood
Instead of querying Aurora compute instances or reading binary logs over database connections, Zero-ETL hooks directly into Aurora's distributed storage tier. As transactions commit, storage log records stream directly to the Redshift cluster's managed storage layer.
Redshift applies these changes continuously, keeping the target database synchronized with sub-minute latency without consuming Aurora CPU or memory.
Architectural trade-offs
- Read-only target: Tables replicated via Zero-ETL inside Redshift are read-only. You cannot execute manual
UPDATEorDELETEstatements on them. - Limited DDL support: Altering or dropping columns with complex types on the source Aurora database can pause or break replication.
- Strict requirements: Source tables must have defined primary keys and use supported collation configurations.
- Cost: Eliminates pipeline engineering overhead, but both source Aurora and target Redshift instances must be sized to handle continuous ingestion apply rates.