Data migration

Data moves separately from the servers that used to hold it, and by a different tool depending on what shape it is in.

AWS DMS handles databases — relational, data warehouses, and several NoSQL and object sources — including heterogeneous moves where the engine changes on the way. It can keep replicating changes right up to the cutover, which is what makes a near-zero-downtime database migration possible.

AWS DataSync handles files and objects: NFS and SMB shares, HDFS, S3-compatible object stores and other clouds, into S3, EFS and the FSx family. It runs one-off migrations and scheduled recurring transfers equally well. Where DataSync ends and S3 File Gateway begins is a common confusion, so there is a page comparing the two.

Two pages here cover the destination architecture rather than the move itself: data lake versus Lake Formation and where ETL and ELT sit in a pipeline.

For the hands-on Oracle and PostgreSQL migration runbooks — CDC checkpoints, recovery procedures, time-windowed reloads — see the DMS section.