AWS Database Migration Service (DMS): Step-by-Step Guide

What is AWS Database Migration Service?

AWS Database Migration Service (AWS DMS) is a managed service for moving data between databases: relational engines, data warehouses, NoSQL stores and services such as Amazon S3. At a basic level, AWS describes it as a server in the AWS Cloud that runs replication software. You tell it where to read (the source endpoint) and where to write (the target endpoint), then run a task on that server to move the data. One of the two endpoints must be on an AWS service; AWS DMS cannot migrate from one on-premises database to another.

AWS DMS handles both homogeneous migrations (SQL Server to SQL Server, PostgreSQL to PostgreSQL) and heterogeneous ones (SQL Server or Oracle to PostgreSQL). For heterogeneous moves the schema has to be converted first. AWS now points to DMS Schema Conversion, a web-based feature inside the DMS console, and calls the downloadable AWS Schema Conversion Tool (AWS SCT) the legacy option. DMS Fleet Advisor, the old discovery feature, reached end of support on 20 May 2026.

If you are deciding where the target database should live, start with AWS database services compared and what Amazon RDS is. For the whole migration process, not just the AWS tooling, see the database migration guide.

How AWS DMS works: replication instance, endpoints and tasks

Replication instance. DMS creates it on an Amazon EC2 instance in your VPC. It connects to the source, reads and reformats the data, and loads the target; most work happens in memory, with large transactions, cached changes and logs spilling to disk. Instance classes are T3 (burstable, suited to testing and periodic loads), C5/C6i/C7i (compute optimised) and R5/R6i/R7i (memory optimised). AWS suggests the R classes for heavy CDC workloads with many transactions per second. A Multi-AZ option keeps a synchronous standby in another Availability Zone, and the console guidance is to turn it on if you intend to use CDC.

Endpoints. An endpoint holds the connection details for one data store: engine, server name, port, credentials (typed in, or held in AWS Secrets Manager) and SSL mode. Extra driver behaviour is set with endpoint settings.

Tasks. A task links a replication instance with a source and target endpoint, selects tables with table-mapping rules and picks a migration type. Each task can also enable data validation, set LOB handling and choose a target table preparation mode.

Step by step: SQL Server to Amazon RDS for PostgreSQL

The steps below follow the order AWS gives in its getting-started tutorial (convert the schema, create a replication instance, create endpoints, create and start a task, verify) applied to a common case: a self-managed SQL Server database moving to Amazon RDS for PostgreSQL with full load plus CDC.

1. Check the source supports CDC

AWS DMS lists SQL Server 2012 to 2022 as sources (2008 and 2008 R2 on DMS 3.5.4). Enterprise, Standard, Workgroup, Developer and Web editions support full load; CDC needs Enterprise, Standard (2016 and later) or Developer. Express edition is not supported. The database must use the Full or Bulk-logged recovery model and have a full backup.

On a self-managed SQL Server, DMS captures changes with MS-Replication for tables with primary keys and MS-CDC for tables without them. With the sysadmin role, the documented MS-CDC set-up is shown below. On Amazon RDS for SQL Server, use exec msdb.dbo.rds_cdc_enable_db 'DB_name' instead of sp_cdc_enable_db. Without MS-Replication or MS-CDC, tables without a primary key only have INSERT and DELETE events captured.

USE [mydb];
EXEC sys.sp_cdc_enable_db;

-- repeat for each table that needs MS-CDC
EXEC sys.sp_cdc_enable_table
    @source_schema = N'dbo',
    @source_name   = N'orders',
    @role_name     = NULL;

2. Convert the schema with DMS Schema Conversion

DMS Schema Conversion uses three resources: an instance profile (network and encryption settings), a data provider for each database (type, server name, port) and a migration project that ties them together with Secrets Manager secrets for credentials. In the project you run an assessment report, convert tables, views, procedures and functions, review the result, and then apply the converted code to the target or export it as SQL scripts to S3. Where the rules cannot convert an object, it can use generative AI to convert more of it for you to review, and it adds an extension pack to emulate some SQL Server features.

Two SQL Server to PostgreSQL settings are worth deciding early. Object names are folded to lower case by default (set case sensitivity on if your code relies on mixed-case names), and the target schema name defaults to database_schema, for example AdventureWorks_dbo; choose the SCHEMA option if you want plain dbo. The type and T-SQL differences you will meet are covered in the SQL Server to PostgreSQL migration guide.

3. Create the replication instance

1. In the AWS DMS console, choose Replication instances, then Create replication instance.
2. Enter a Name, choose an Instance class and an Engine version.
3. Under High Availability, choose the Multi-AZ option if the task will run CDC.
4. Set Allocated storage (GiB); AWS says the default is usually enough but to watch storage metrics during large loads.
5. Choose the VPC and Replication Subnet Group (the one containing the source if it is in a VPC), and decide on Publicly accessible, which cannot be changed later.
6. Under Advanced, pick the Availability Zone of the source, the VPC security group and the KMS key.

4. Create the source and target endpoints

Choose Endpoints, then Create endpoint, once for each side. Set Endpoint type (source or target), the engine, an Endpoint identifier, and either Secrets Manager or the server name, port, user name and password. For an RDS target you can pick Select RDS DB Instance instead of typing the host. The DMS user on SQL Server needs the permissions listed in the SQL Server source chapter; sysadmin is the simplest route on a self-managed server.

5. Create the task and start it

1. Choose Database migration tasks, then Create task.
2. Pick the replication instance and both endpoints, and set Migration type to Migrate existing data and replicate ongoing changes.
3. Because the schema was created in step 2, set Target table preparation mode to Do nothing or Truncate. Drop tables on target recreates tables with primary keys only, without secondary indexes, other constraints or column defaults.
4. Choose LOB handling: Limited LOB mode is faster but truncates values above the maximum LOB size; Full LOB mode copies everything more slowly.
5. Tick Enable validation, add a selection rule for the schema, and choose Create task.

On PostgreSQL, foreign keys are implemented with triggers, and AWS strongly recommends disabling them during the full load, for example with the session_replication_role parameter. AWS also recommends creating secondary indexes after a large full load rather than before, unless the data set is small.

{
  "rules": [
    {
      "rule-type": "selection",
      "rule-id": "1",
      "rule-name": "1",
      "object-locator": { "schema-name": "dbo", "table-name": "%" },
      "rule-action": "include"
    }
  ]
}

aws dms create-replication-instance \
  --replication-instance-identifier sqlserver-to-pg \
  --replication-instance-class dms.t3.medium \
  --allocated-storage 100 \
  --multi-az

aws dms create-endpoint --endpoint-identifier sqlserver-source \
  --endpoint-type source --engine-name sqlserver \
  --server-name <source-host> --port 1433 \
  --username <dms-user> --password <your-password> --database-name mydb

aws dms create-endpoint --endpoint-identifier postgres-target \
  --endpoint-type target --engine-name postgres \
  --server-name <rds-endpoint> --port 5432 \
  --username <pg-user> --password <your-password> --database-name mydb

aws dms create-replication-task \
  --replication-task-identifier mydb-full-load-cdc \
  --source-endpoint-arn <source-endpoint-arn> \
  --target-endpoint-arn <target-endpoint-arn> \
  --replication-instance-arn <replication-instance-arn> \
  --migration-type full-load-and-cdc \
  --table-mappings file://table-mappings.json

aws dms start-replication-task \
  --replication-task-arn <task-arn> \
  --start-replication-task-type start-replication

6. Validate and cut over

With validation enabled, DMS starts comparing rows as soon as each table finishes its full load and then checks CDC changes as they arrive; validation needs a primary key or unique index on each table. Before cutover, stop application writes, let the remaining changes apply, then stop the task. DMS does not migrate sequences during ongoing replication, so set each PostgreSQL sequence (or identity) to the next value after replication stops, then point the application at the new endpoint. Our database migration checklist lists the checks to run before and after the switch.

AWS DMS Serverless and homogeneous data migrations

DMS Serverless removes the replication instance. You still create endpoints, but instead of a task you create a replication configuration; when it starts, DMS reads the source metadata, works out the capacity it needs and scales between the minimum and maximum DMS capacity units (DCUs) you allow. One DCU equals 2 GB of RAM, and AWS offers 1, 2, 4, 8, 16, 32, 64, 128, 192, 256 and 384 DCUs. The current Serverless engine version is 3.5.4, and it has its own list of limitations to check before you choose it.

Homogeneous data migrations are a separate, instance-free option for like-to-like moves (MySQL or MariaDB, PostgreSQL, and MongoDB or Amazon DocumentDB) that uses the engines' native tools, and is billed per migration hour.

Azure Database Migration Service in brief

Azure Database Migration Service (Azure DMS) is Microsoft's managed migration service. Its current version is driven from the Azure portal, PowerShell and the Azure CLI, and the overview page says it currently supports SQL Server modernisation to Azure: offline migrations to Azure SQL Database, and online or offline migrations to Azure SQL Managed Instance and SQL Server on Azure Virtual Machines, including schemas. Assessment, SKU recommendations, login migration and TDE-encrypted databases are not part of the portal version (separate PowerShell and CLI tutorials cover logins and TDE). Microsoft also documents MySQL, PostgreSQL and MongoDB migration tutorials under the same service.

The Azure pricing page lists a free Standard tier for offline migrations and a paid Premium tier for online migrations. If your target is Azure SQL Database or Azure SQL Managed Instance, compare Azure DMS with SQL Server Migration Assistant, which converts Oracle, MySQL, Access, Db2 and SAP ASE schemas to SQL Server.

When to use AWS DMS, and alternatives

• Good fit: moving a database into AWS with little downtime, changing engines (with Schema Conversion first), consolidating several sources, or keeping a copy in sync with CDC for a period.
• Poor fit: small databases you can stop for an hour (a native dump and restore is simpler, see database migration methods), on-premises to on-premises moves (not supported), and schema-only work.
• Limits to plan for: sequences, secondary indexes, users and code objects are not carried by the data task; LOB settings trade speed against truncation; validation needs keys.
• Alternatives: native backup and restore or dump tools for same-engine moves, pgloader for MySQL or SQL Server to PostgreSQL, and Azure DMS or SSMA when the target is Microsoft's platform.

Pricing

Prices from the provider's official pricing pages, checked 8 October 2026, region US East (N. Virginia), in USD, excluding tax. List prices only; discounts, commitments and your actual usage change the bill.

AWS DMS has three pricing options: on-demand replication instances, DMS Serverless and Database Savings Plans. On-demand instances are billed per hour by instance class, with a separate rate for Multi-AZ. T3 instances run in unlimited mode, so sustained CPU above the baseline is charged as CPU credits at 0.075 USD per vCPU-hour. T2 and T3 instances include 50 GB of storage (C4, R4 and R5 include 100 GB); extra replication storage is billed per GB-month. DMS Serverless is billed per DCU-hour, homogeneous data migrations per migration hour, and DMS Schema Conversion is free apart from Amazon S3 storage.

Data transfer into DMS is free, as is transfer between DMS and RDS or EC2 databases in the same Availability Zone; standard AWS rates apply across AZs, Regions or out of AWS. Free tier: accounts that signed up before 15 July 2025 get 750 hours a month of Single-AZ dms.t3.micro and 50 GB of storage for a year; accounts created after that date get a credit-based Free Plan or Paid Plan with 100 USD in credits (and up to 100 USD more), valid for up to twelve months. Prices exclude tax.

Worked examples are estimates calculated from the list prices above; they are not quotes or measured bills.

Frequently asked questions

Sources

• AWS DMS User Guide: What is AWS Database Migration Service?
• AWS DMS User Guide: Getting started
• AWS DMS User Guide: Working with a replication instance
• AWS DMS User Guide: Choosing a replication instance type
• AWS DMS User Guide: Creating a replication instance
• AWS DMS User Guide: Working with endpoints
• AWS DMS User Guide: Creating source and target endpoints
• AWS DMS User Guide: Creating a task
• AWS DMS User Guide: Selection rules (table mapping)
• AWS DMS User Guide: Sources for data migration
• AWS DMS User Guide: Using SQL Server as a source
• AWS DMS User Guide: CDC for SQL Server as a source
• AWS DMS User Guide: Using PostgreSQL as a target
• AWS DMS User Guide: Best practices
• AWS DMS User Guide: Data validation
• AWS DMS User Guide: Converting schemas with DMS Schema Conversion
• AWS DMS User Guide: SQL Server to PostgreSQL conversion settings
• AWS DMS User Guide: Working with AWS DMS Serverless
• AWS DMS product page
• AWS DMS pricing
• AWS CLI reference: create-replication-instance
• AWS CLI reference: create-endpoint
• AWS CLI reference: create-replication-task
• AWS CLI reference: start-replication-task
• AWS price data: DMS replication instances and storage
• AWS price data: DMS Serverless and homogeneous migrations
• Microsoft Learn: What is Azure Database Migration Service?
• Azure Database Migration Service pricing

Checked 8 October 2026.

How we research cloud database guides: our editorial method. CodeWithSQL earns nothing from the providers mentioned.