AWS Application Migration Service
A service for lift-and-shift migration of on-premises or other cloud servers to AWS, minimizing cutover downtime through continuous replication
Overview
AWS Application Migration Service (MGN) is a service for migrating physical servers, virtual machines, and other cloud instances to AWS as EC2 instances. Simply installing an agent on the source server initiates block-level continuous replication, enabling migration to AWS without any changes to the OS or applications. The two-stage process of test launch and cutover allows safe pre-production verification before the actual switchover.
Agent-Based Block Replication Enabling Near-Zero Downtime Migration
The MGN migration process begins with installing the AWS Replication Agent on the source server. The agent detects changes at the disk block level and continuously replicates them to a staging area on the AWS side (lightweight EC2 instances and EBS volumes). After the initial full sync completes, only deltas are transferred, limiting the impact on network bandwidth. This continuous replication keeps the source server running normally while maintaining an up-to-date replica on AWS at all times. At cutover, you simply launch production EC2 instances from the replica and switch DNS or load balancers, keeping downtime to just a few minutes for reboot. Azure Migrate offers similar agent-based migration, but MGN has lower replication latency and excels at parallel processing in large-scale migration waves.
Two-Stage Process of Test Launch and Cutover
To ensure migration reliability, MGN clearly separates the Test and Cutover stages. During test launch, EC2 instances are started from replicated data to perform application verification, network connectivity testing, and performance validation. Test instances launch in an environment independent of production, so they do not affect source server operations. Once testing succeeds, you execute the cutover to launch production instances. Pre-defining instance type, subnet, security group, and IAM role in launch templates eliminates manual configuration during cutover. Cloud migration books on Amazon cover practical processes from migration wave planning through cutover. If fallback to the source server becomes necessary after migration, you can configure replication in the reverse direction.
Large-Scale Migration Wave Planning and Migration Hub Integration
For migrations involving hundreds of servers, it is essential to divide migration targets into waves, grouping servers with dependencies into the same wave. MGN supports grouping servers by wave and application concepts, enabling batch test launches and cutovers at the wave level. Integration with Migration Hub automatically reflects MGN migration progress on the Migration Hub dashboard, providing a unified view of overall progress alongside DMS database migrations. In the pre-migration discovery phase, Migration Hub's Application Discovery Service or Agentless Collector collects server configurations, dependencies, and performance data from the on-premises environment as foundational data for migration planning. MGN itself is free to use, with charges applying only to the destination EC2 instances and staging area EBS volumes. The staging area uses low-cost t3.small instances by default, keeping costs minimal during the replication period.