Stream CDC into an S3 Data Lake in Parquet with AWS DMS

Part 8: Hands-on Tutorials · Hands-on Tutorials

Read the full tutorial: Stream CDC into an S3 Data Lake in Parquet with AWS DMS ↗

Link verified 2026-08-23.

This tutorial points AWS DMS at a source relational database, replicating both a full initial load and every subsequent insert/update/delete as change data capture events, and lands each as Parquet files in S3.

It’s the concrete, single-source version of the ingestion decision covered in Ingestion Decisions: Batch, Incremental Loading & Change Data Capture — DMS here plays the role of the log-based CDC connector that topic argues for over trigger- or query-based polling.

flowchart LR
    A[(Source RDBMS)] --> B[AWS DMS replication instance]
    B -->|Full load| C[S3 Parquet - initial snapshot]
    B -->|CDC stream| D[S3 Parquet - ongoing changes]
    C --> E[Data lake / catalog]
    D --> E

---

| ← Previous: Serverless Data Lake Analytics with AWS Glue, S3 & Athena | Next: Transactional Data Lake with Apache Iceberg, EMR Serverless & Athena → | |:—|—:|