A year ago Matthew Dillon went into some detail about the new HAMMER filesystem. It’s now reaching stability, and will hopefully be sent forth in the upcoming v2.0 release of DragonflyBSD. All the interesting bits can be discovered here and here. A few features to note, though, are…
1 Exabyte of storage
Online data migration, replication and evacuation – multi-master disk replication across WANs anyone?
Logical data retention policies – ie you can have block level replication to a remote site with large slow disks, but with a lengthy data retention policy
Inbuilt volume level and file level versioning and snapshotting – just mount up or request a version number, and you get a snapshot from that time. A bit like MS VSS.
Inbuilt reblocker for defragging, expansion and contraction.
PF linked connection state recovery – keep those TCP links alive throughout a router reboot.
These are just a few of the features, read the documentation to find out the rest. It may turn out to be a ZFS killer, but anything with replication like this sounds good to me. Now we just need to wait for it to be ported to Linux, assuming Zumastor and Tux3 don’t rule the roost by then 🙂 For what it’s worth, there’s some info on Tux3 and HAMMER features and ideas at http://kerneltrap.org/Linux/Comparing_HAMMER_And_Tux3