We know that a one-time full load of existing data is comparatively simple to set up and run, but many production applications with large databases cannot tolerate a downtime window long enough to migrate all the data in a full load. AWS DMS can leverage a proprietary Change Data Capture (CDC) process to conduct ongoing replication from the source database to the target database for these databases. AWS DMS manages and monitors the ongoing replication process with minimal load on the source database without platform-specific technologies and components which need to be installed on either the source or target.

CDC offers two ways to implement ongoing replication:

• Migrate existing data and replicate the changes - implements ongoing replication by:
  • (Optional) Creating the target schema
  • Migrating existing data and caching changes to existing data as it is migrated
  • Applying those cached data changes until the database reaches a normal state
  • Lastly, applying current data changes to the target as soon as they are received by the replication instance
• Replicate data changes only (no schema) from a specified point in time. This option is helpful when the target schema already exists and the initial data load is already completed. For example, using native export/import tools, ETL, or snapshots might be a more efficient method of loading the bulk data in some situations. In this case, AWS DMS can be used to replicate changes from when the bulk load process started to bring and keep the source and target databases in sync.

AWS DMS takes advantage of built-in functionality of the source database platform to implement the proprietary CDC process on the replication instance. This allows AWS DMS to manage, process, and monitor data replication with minimal impact to either the source or target databases.