Large-scale data-center migration with AWS MGN and DataSync
One data center, two data problems: running servers and the file estate underneath them. MGN and DataSync each own one half.
Two migration tools, two kinds of data
A large data-center exit is really two migrations happening at once. There are the running servers (application tiers, databases, middleware) whose whole disk and live state you want to lift to AWS as-is, and there is the file estate (departmental shares, home directories, media libraries, archives) sitting on NAS and file servers as flat NFS and SMB data. Trying to move both with one tool is the classic wrong answer.
AWS Application Migration Service (MGN) owns the servers. It is the recommended rehost (lift-and-shift) service: it automates migrating physical, virtual, and cloud servers to AWS with minimal downtime by converting them for launch as EC2 instances. AWS DataSync owns the bulk file and object data: it is a secure, high-speed transfer service purpose-built for moving file or object data to, from, and between AWS storage services. Pick the tool by what you are moving, not by the destination.
Two migrations at once: match the tool to the data
Trace
A data-center exit is really two migrations. The running servers (application, database, middleware) carry live whole-disk state and belong to AWS MGN, which rehosts them as EC2 instances. The flat file estate (NAS shares, home directories, media, archives) is loose NFS/SMB data and belongs to AWS DataSync, which lands it on S3, EFS, or FSx. Pick the tool by what you are moving, not by the destination.
How MGN rehosts a server
You install a lightweight MGN replication agent on each source server. The agent performs continuous block-level replication of the server's disks over the network into AWS, so the copy stays current as the source keeps running. That replicated data lands in a staging area subnet on lightweight EC2 replication servers that write it to EBS volumes; this staging footprint is cheap because it is not running your production instances yet.
Because replication is continuous, you can run non-disruptive test launches from the current point-in-time without touching the source, validate the application in AWS, then keep replicating until you are ready. The final cutover launches the production EC2 instances and typically takes only a cutover window of minutes, which is what keeps the RTO low. To manage scale you group servers into applications and applications into waves, so launch, test, and cutover can be driven as bulk operations across hundreds of servers rather than one at a time.
How MGN rehosts a server: replicate, test, cut over
Trace
A lightweight MGN agent on each source server streams continuous block-level replication into a cheap staging-area subnet, where lightweight replication servers write the data to EBS. Because the copy is always current, you run non-disruptive test launches from a point-in-time, then cut over by launching the production EC2 fleet in a window of minutes. Servers are grouped into applications and waves so launch/test/cutover run as bulk operations across hundreds of hosts.
How DataSync moves the file estate
For the file and object data, you deploy a DataSync agent (typically an on-premises VM) that reads the source shares. DataSync moves data from on-premises NFS, SMB, HDFS, and object storage into AWS targets including Amazon S3, Amazon EFS, and Amazon FSx. It uses a purpose-built, parallel, multi-threaded protocol to move data fast, and it handles incremental transfers so recurring syncs only move what changed.
Every transfer includes automatic encryption in transit and end-to-end data-integrity validation, so the data arrives securely and intact. DataSync accesses AWS storage through IAM roles and can use VPC endpoints, which lets the transfer avoid traversing the public internet entirely. That combination (validation plus private-path option) is why DataSync, not a hand-rolled rsync or copy script, is the exam-correct answer for large, one-time or repeated file migrations.
How DataSync moves the file estate — securely and privately
Trace
An on-premises DataSync agent reads the NFS, SMB, HDFS, and object sources and moves them with a purpose-built parallel protocol. Every transfer is automatically encrypted in transit and end-to-end integrity-validated. DataSync reaches AWS storage through IAM roles and, via a VPC endpoint, keeps the whole transfer off the public internet, landing data on S3, EFS, or FSx. Incremental syncs move only what changed.
The network path and the cutover choreography
Both data streams share a private, high-bandwidth path into AWS: AWS Direct Connect is the primary link, with Site-to-Site VPN as a resilient backup. Keeping MGN block replication and DataSync file copy on the same private path keeps bulk migration traffic off the public internet and predictable in cost.
The choreography matters for the exam. MGN replication runs continuously in the background while the business keeps operating, so cutover downtime is measured in minutes, not the full copy time. DataSync's incremental syncs let you pre-seed the file estate early and run a final delta sync at cutover, so users see current data on day one. The tools do not overlap: never point MGN at a plain file share (it rehosts whole servers, not loose files), and never try to lift a running server with DataSync (it copies files, not live block state). Match the tool to the data and the two halves of the migration cut over cleanly.
Shared private path and the cutover choreography
Trace
Both migration streams — MGN block replication and DataSync file copy — share one private, high-bandwidth path: Direct Connect primary, Site-to-Site VPN as resilient backup, keeping bulk traffic off the public internet. MGN replicates continuously in the background so cutover is minutes, not the full copy time; DataSync pre-seeds the file estate early and runs a final delta sync at cutover, so users see current servers and current files on day one.
