Btrfs Filesystem In Linux 3.6 Kernel Has Big Changes

[Markore]

Originally Posted by Markore View Post
ZFS is only filesystem you can count on to be supported on all platforms (Windows remain to be ported to

Fat32 and UDF?

I was thinking of Filesystem you can install your operating system to.
And from what you can boot multiple different operating systems,
and using same boot/system disk(s) together with shared data disk(s) pools.

ZFS, being supported for boot from Illumos,Linux and FreeBSD is the best system I know for the moment i you want to be free of OS/Platform constraints or even architecture constraints (same ZFS pool works on SPARC and x86 without changes)

I do not believe Btrfs for stability and production use for now.
If I tend to be supported with Linux install I would go with ext4 for system disks (I think ext4 now has snapshots)
and ZFS for data pool that applications actually use to work with or store data to.

[tstrunk]

ZFS is still great and for many things the only choice still.
But it will quite certainly never be a native (in-kernel and distribution default) filesystem for linux, due to licensing and patenting.

Btrfs didn't reach feature-parity yet, but the developers promise that it will. And it will be native and not patent-encumbered. We need a filesystem like that on Linux, because ext4 doesn't have all the advanced features (especially checksums in the metadata). Therefore it's very good for the health of the linux ecosystem that it exists.

Of course. ZFS is used on Linux for LUSTRE cluster file system and most people that used to develpo it for Sun are now in companies outside of Oracle, developing it further.
CERN use ZFSOnLinux for its physics measurements that could not have any errors
and Zfs on Linux is used to make petabyte storage (PB).
- If it is good enough for science laboratories, it is good enough for me.

As far as I (and wikipedia) know, Lustre is still quite independent from ZFS:
"The storage used for the MDT and OST backing filesystems is partitioned, optionally organized with logical volume management (LVM) and/or RAID, and normally formatted as ext4 file systems. The Lustre OSS and MDS servers read, write, and modify data in the format imposed by these file systems." (from wikipedia)
It does feature a ZFS storage backend though.

[Kano]

Maybe you should read:

http://www.phoronix.com/scan.php?pag...tem&px=MTExMTc

[tstrunk]

Maybe you should read:

http://www.phoronix.com/scan.php?pag...tem&px=MTExMTc

That's only for Metadata. It's very fast as written in the article, but it cannot guarantee that your data is not corrupted.

[devsk]

The beauty of ZoL is that it installs as a separate package and is modular. So, putting it inside of kernel is not needed. There is no patching of the kernel (unlike R4).

How hard is it to install a package and use the filesystem? Without breaking any licensing/legal laws.

ZoL is the best thing that happened to Linux in recent times (like last 2 years)...