Linux will soon have btrfs, which is as good as (if not better than) ZFS.
Also Dtrace & ZFS originally come from Solaries, so they are not FreeBSD projects. The only reason they do not work on Linux is because of license problems which Sun created.
>Linux will soon have btrfs, which is as good as (if not better than) ZFS.
btrfs does not even try to solve the problem I want zfs for; snapshots over the network.
You see, when you deal with a large amount of storage (especially if you deal with that storage as block devices that run arbitrary filesystems, and not just files on a filesystem) the bottleneck for backups becomes disk bandwidth, not network bandwidth, so things like rsync don't help much.
ZFS snapshots over the network? those help rather a lot. Unlike rsync and stuff, it saves me disk bandwidth, not just network bandwidth. Unlike inotify-based systems, I don't need to have knowledge of the filesystems I'm replicating.
(I know I harp on this a lot... I just want to point out that btrfs, no matter how good it is for the problems it attempts to solve, isn't even trying to solve the problem I need ZFS for.)
Ooh. thanks. this post is probably the most useful bit of information I've read this month. Hm. I will have to research more, but this might... change things significantly. I mean, it's still a newish feature, not something you want for backups, but between that an bcache, Linux looks like it might deserve another look, storage-wise.
At <company I work for, which owns pfSense> we use ZFS for exactly this. Functional, always 'on' backups to the hosting center next door (we have some nice single-mode fiber running to our cabinet there.)
> Linux will soon have btrfs, which is as good as (if not better than) ZFS.
It may be awesome, but I'd say "soon" is a pretty optimistic timeline.
Our testing pretty conclusively demonstrated that btrfs was still a long way from production-ready; We've seen numerous problems with deadlocks; Filling up a filesystem pretty much ruins it forever -- And these are not new problems.
As far as I can tell, those forecasts have almost always come from outside the project. I don't recall Chris Mason ever saying that it's right around the corner.
In fact, searching for "soon" on the Btrfs wiki produces 0 results.
It's still a very active project, but there's still a lot of work to be done.
At this point, the amount of complete data losses I've seen on the LKML makes me wonder why anyone wants to promote BTRFS as being better than anything I would want to use for things approaching production.
I may be an old Solaris hand now but ZFS works. Now, not in the future, my home fileservers on FreeBSD have worked for ages and had no issues. Unlike the many complete losses on BTRFS on Linux that I see happening time and time again. SuSE might be promoting BTRFS, but objectively I can't agree with their decision.
One thing I like about FreeBSD, and the BSDs in general, is that they don't suffer from such problems.
It's odd that Oracle (who now own Sun's IP) paid to develop a whole new filesystem (btrfs) for Linux instead of just changing the licensing terms for ZFS. It would be nice to have a really good, Apache-licensed filesystem that could run across Linux, OS X, BSD, and Solaris.
Also Dtrace & ZFS originally come from Solaries, so they are not FreeBSD projects. The only reason they do not work on Linux is because of license problems which Sun created.