AWS Database Migration Service

Migration & Transfer

Managed service that moves databases into AWS and keeps them in sync until you cut over.

Verified against AWS documentation · 2026-09-09

Key facts

  1. 01The source database stays online and serving traffic — you cut over only once the target has caught up
  2. 02Same engine (MySQL → RDS MySQL) is a straight data move; different engines (Oracle → Aurora PostgreSQL) need schema conversion first, via DMS Schema Conversion or the downloadable AWS SCT
  3. 03Three task types: full load (one-time copy), CDC (ongoing changes only), full load + CDC (copy, then stay in sync)
  4. 04Runs on a replication instance you size and manage, or DMS Serverless which picks and scales capacity for you
  5. 05Reality check: DMS Fleet Advisor, the fleet discovery feature, hit end of support on 20 May 2026 — pre-2026 study material still lists it

Exam triggers

Phrases and scenarios that point here.

  • "Migrate a database to AWS" + "minimal downtime" → DMS
  • Source and target engines differ → convert the schema first (DMS SC or AWS SCT), then DMS moves the data
  • "Keep both databases in sync during cutover" / phased migration → DMS with CDC
  • Continuously replicate production tables into Redshift or S3 for analytics → DMS, not a nightly export script

Not this service

  • Moving whole servers or VMs, not just their data → AWS Application Migration Service (MGN)
  • Copying files, file shares, or S3 objects → AWS DataSync
  • Petabytes with no network that can carry them → AWS Snow Family
  • Tracking overall migration progress across teams and tools → AWS Migration Hub

Use cases

  1. 01Weekend cutover of a self-managed MySQL on EC2 into RDS — full load runs Friday night, CDC keeps it current, you swap the connection string Sunday
  2. 02Escaping commercial database licensing: SCT converts the Oracle schema to Aurora PostgreSQL, DMS moves the rows
  3. 03Standing up a reporting copy — replicate live tables into Redshift or S3 so analysts stop hammering the transactional database
  4. 04Data centre exit: consolidating a dozen small on-prem databases into a handful of RDS instances without rewriting the apps
  5. 05Running old and new systems side by side during a phased rewrite, with CDC keeping both databases current until the last service moves

How it's charged

  1. 01On-demand: hourly rate for a replication instance, priced like EC2 — you pay for time and capacity, so a slow migration costs more than a big one
  2. 02Serverless: per DCU-hour (1 DCU = 2 GB RAM), capacity scales itself; homogeneous migrations bill hourly with no instance and no storage charge
  3. 03Schema conversion is free — you only pay for the S3 storage holding assessment reports and converted code
  4. 04Data into DMS is free, and DMS ↔ RDS/EC2 traffic in the same AZ is free; cross-AZ, cross-Region, or out of AWS bills at standard transfer rates
  5. 05Billing trap: nothing stops a replication instance after cutover — a forgotten CDC task bills around the clock

Commonly confused with

AWS Application Migration Service (MGN)Migration & Transfer

References