Gee that's tough.

Posted Mar 20, 2007 4:47 UTC (Tue) by drag (guest, #31333)
In reply to: Gee that's tough. by qu1j0t3
Parent article: The 2007 Linux Storage and File Systems Workshop

Most of the features of ZFS are aviable on Linux right now.

LVM, MD, DM, and lots of other FS-related features can be combined in different ways that will accomplish do the vast majority of what ZFS does. All of it's been around for a long time and is proven. It's just in many smaller components instead of one big monolythic package.

The main thing that ZFS brings to the table is that it's simplier to administrate. There are a few fancier features that Linux doesn't have, like 128bit-ness, but there are a lot of things that Linux can do that ZFS can't.

Essentially if Linux developers decided to adopt ZFS they'd be replicating existing functionality.

Plus ZFS is copyrighted to Sun and is not aviable for Linux inclusion due to licensing differences.

Also I expect that Sun has patents on various aspects of ZFS, so unless Linux is able to incorporate code from Sun into the kernel then it's likely that Linux developers would violate obvious patents if they tried to re-impliment it.

Just my perspective on the whole 'zfs' issue.

to post comments

ZFS

Posted Mar 20, 2007 8:06 UTC (Tue) by job (guest, #670) [Link]

I don't think that is the main thing ZFS brings. Of course, for Solaris admins, the main thing is that ZFS brings logical volume-type functionality which was badly missing from vanilla Solaris before.

But to me from a Linux background, it brings 1) much better snapshot capabilities (more of them with better performance) due to it being copy-on-write at the file system level, 2) checksummed data integrity (I haven't seen this elsewhere but I suspect this is great for running low end storage systems such as software RAID) and 3) better performing software raid-5 ("raid-z") because of its knowledge how the files are laid out.

I like ZFS. I haven't used it in production (and probably won't, at my current work), but it is a solid piece of engineering. It seems a bit bastardized at first as it is both a file system and a volume manager, but I think it is well motivated.

Gee that's tough.

Posted Mar 20, 2007 15:18 UTC (Tue) by cajal (guest, #4167) [Link]

"LVM, MD, DM, and lots of other FS-related features can be combined in different ways that will accomplish do the vast majority of what ZFS does."

No, they can't. They don't give you self-repairing storage with provable data correctness. They don't give you dynamic striping. The LVM still requires you to manually carve up PVs into LVs. ext3 doesn't support arbitrary number of extended attributes. I could go on.

Further, just because the GPL is not compatible with Sun's implementation of ZFS, is no reason that the Linux kernel community couldn't reimplement ZFS. The on-disk format and man pages are available at http://opensolaris.org/os/community/zfs/docs/ I think this could be a valuable piece of software for the Linux kernel.

Gee that's tough.

Posted Mar 21, 2007 13:40 UTC (Wed) by mennucc1 (guest, #14730) [Link] (2 responses)

> Most of the features of ZFS are aviable on Linux right now.

but still I would like to have snapshots in EXT (maybe in 5 :-) ?)

Gee that's tough.

Posted Mar 21, 2007 16:06 UTC (Wed) by bronson (guest, #4806) [Link] (1 responses)

For snapshotting, just run whatever filesystem you want on top of LVM. LVM is a little hard to get used to at first but it's definitely worth the effort. Here are some notes I took when setting it up on my systems: http://wiki.u32.net/LVM

I now put all nontrivial partitions on LVM. Works for me. I'll let others argue whether LVM snapshots are worse than ZFS, or if ZFS is a layering violation. :)

Gee that's tough.

Posted Mar 22, 2007 5:55 UTC (Thu) by snitm (guest, #4031) [Link]

LVM2 Snapshots are quite bad. For starters they are done at the block-level whereas ZFS provides file-level snapshots (aka redirect on write). LVM2 snapshots don't scale well either; seeing as each snapshot imposes a copy out penalty because there isn't a shared exception store (aka LVM snapshot LV) for all snapshots of an origin LV.