Replication is a ZFS mechanism to transfer filesystems under a Solaris derivate with all properties like ACL or real Windows SIDs between pools either locally or over the network. Windows user permissions from an Active Directory as a file attribut on ZFS stay intact on replications (only with Solaris CIFS server, not when using SAMBA). Replication is therefor not only more powerful than for example rsync but also much faster. Napp-it supports high performance replication over buffered netcat connections. Performance is near to physical limits (ex network limits).
A Replication starts with a snapshot of the source filesystem. This snapshot is transferred. The following replications are then incremental based on delta-snaps. They transfer modified data only. Replication is very fast and allows to keep two filesystems in sync down to a minute even under heavy load.
You should only consider that the state of a snapshot is similar to a sudden power-off. If you want to replicate over unsafe networks like the Internet you should use a VPN tunnel.
Replication is the first method for availability and backup. You can keep adjustable snapshots on the backupsystem for versioning and read-only access to former data states. Easiest way to access these former states is via CIFS shares and Windows - previous version.
If you want to switch to the replication system if a primary system
fails, you only need to stop the replication job, deactivate the
read-only property of the filesystem and activate/share file services or
replicated blockbased iSCSI zvols/LUNs. To go back to the former state
after repair, either use the repaired system as backupserver or replicate the
filesystem back, rename or delete the filesystem on your backupsystem and restart the old replication job.