On Mon, May 18, 2015 at 2:21 PM, Nuno Magalhães <nunomagalh...@eu.ipp.pt> wrote: > On Mon, May 18, 2015 at 2:08 AM, Rich Freeman <ri...@gentoo.org> wrote: >> On the other hand, both btrfs and zfs will get you a level of data >> security that you simply won't get from ext4+lvm+mdadm - protection >> from silent corruption. > > That's one of the advantages i see in ZFS. Do you use it frequently? > Can anyone comment on its memory usage (without dedicated SSDs for > ARC)? > > Maybe i'd use the 4 drives as a ZFS pool. > > For now LVM complains about duplicate PVs (when pvcreate /dev/md0), > i'll have to fiddle with the filters in lvm.conf >
I use btrfs heavily, but not ZFS. The feature set overlaps, but ZFS has more enterprise-oriented features and is more mature, and btrfs has more single-workstation-oriented features and is less mature. For example, ZFS has features like write-intent logging and read caching that are useful especially on large arrays. Btrfs, on the other hand, lets you mix different-sized drive in a single redundancy unit or add/remove devices to a single redundancy unit, while in ZFS you can have multiple vdevs in a zpool but you cannot add/remove drives from a vdev or fully utilize drives of different sizes in a vdev. That is something which is very useful when you have a 3-drive RAID and want to make it a 4-drive RAID, but it isn't terribly useful when you want to add 5 drives to a 30-drive SAN. I'm not sure how many of those differences are design-limitations vs just being what devs have spent their time on. I'm sure over time the feature set of both will grow and further overlap each other. However, right now with the current focus I'd expect ZFS to continue to focus on features useful in very large deployments, and btrfs on features useful in small deployments. -- Rich