The basic requirement is to allow Labs endusers to recover past versions of files/databases for disaster and error recovery.

This used to work with timetravel snapshots, but that feature was disable when it was suspected to cause instabilities late 2013, and has remained disabled since. The true culprit having been found, reenabling it should be the first layer of redundancy.

In addition, replication of hourly snapshots between the DCs will provide for an cross-site backups.

As this covers only the filesystems, a mechanism needs to be devised by which users can also make safety backups of databases. In practice, a simple mechanism allowing scheduled dumps of the database to the filesystem will allow for the latter to handle redundancy.