ZFS is kind of like raid, lvm, and a filesystem all in one. Instead of using a raid card to combine drives, creating a physical volume/volume group, and the creating filesystems on top of that, everything is controlled by ZFS. It breaks all the abstraction layers, but that turns out not to be a horrible thing when dealing with filesystems.
You can combine 6 disks, for example, in a configuration that can withstand 2 drive failures (similar to raid6, but it's called raidz2). All data is checksummed and auto-healing, so there's no fsck command, although there is a background scrub you should run occasionally. And it uses copy-on-write, so you can snapshot/clone a huge filesystem instantly, and it won't use any additional space until you make changes to one of them (and then, only enough space to store block-level differences). I use this for snapshots - I have snaphots of my home directory every four hours going back two weeks. And new filesystems can be created/expanded/reduced in size almost instantly. This also makes consistent backups really easy.
The only downside is memory consumption. But in a dedicated file server that's not too big a problem.
I'd like to use OpenBSD almost exclusively, but the lack of good virtualization support (no, sparc ldom does *not* count) and lack of a fuse-like filesystem make it a pain. FreeBSD works, and I like ports, but I don't like having to compile everything from scratch all the time, and don't really want to bother setting up my own build server. NetBSD could work, but I never quite got along with pkgsrc for some reason (which is my fault, not pkgsrc's). DragonflyBSD also uses pkgsrc. Debian generally just works,