Slashdot Mirror
Sun Releases ZFS
An anonymous reader writes "Sun's engineers have been blogging today that Sun has finally released its next generation filesystem, ZFS today by pushing out the "community" (i.e. testing) build 27 of OpenSolaris in source and binary form. There is also documentation and a a source code tour available on their site."
About two years ago Solaris seemed doomed. Linux with the advent of 2.6 kernel started to be a real enterprise level contender. Now with ZFS and DTrace Solaris (and Open Source licencse) looks to be a real contender.
You can defy gravity... for a short time
answered my own question... cool.
From OpenSolaris License
3.6. Larger Works. You may create a Larger Work by combining Covered Software with other code not governed by the terms of this License and distribute the Larger Work as a single product. In such a case, You must make sure the requirements of this License are fulfilled for the Covered Software.
Michael J. Ryan - tracker1.info
I cant wait to see it in BSD. Linux would be cool too, but it already has XFS, ReiserFS and other advanced filesystems. I'm not even sure if theyre in the same class of filesystems. I just feel like in a corner with BSD's currently limited selection.
Sometime in the future we'll have one kernel, where we can swap in and out any driver, scheduler, filesystem, device etc from Linux, BSD or Solaris, and compile it as a monolith or micro kernel.
"Give orange me give eat orange me eat orange give me eat orange give me you." -Nim Chimpsky
This is incredibly good news for Sun, and yet another astounding achievement this year.
They open-sourced Solaris (despite the whinging of the nay-sayers and accusations of being in bed with SCO^H^H^HCaldera), they sell Opteron workstations and servers running 64-bit Solaris, 8-core 32-thread Niagara (aka UltraSPARC T1) came out early (the first Sun processor to do so this decade) and now they've pushed out ZFS - the best filesystem ever devised.
If only they can get Project Janus integrated and out in the open...
Stick Men
Watch a demo of it here: http://www.opensolaris.org/os/community/zfs/demos/ basics/
They create 100 file systems in 20 seconds! Amazing!
I'm impressed. Some real innovation in filesystems, and coming from a company some considered to be sinking.
/usr, /home, you know the drill), and each one will use as much space as it actually needs. I suppose you can still have / read-only, /home read-write, etc.
... full 64-bit file offsets''
/smug-lisp-weenie-mode
I love:
``ZFS presents a pooled storage model that completely eliminates the concept of volumes and the associated problems of partitions''
As far as I understand it, there is no need to decide in advance how large your filesystems are going to be. Simply make all your disks one large ZFS pool, then create your filesystems (/,
``All operations are copy-on-write transactions, so the on-disk state is always valid.''
And that seems to go not just for the directory structure, but for the file contents, too.
``ZFS provides unlimited constant-time snapshots and clones.''
Another killer. Clones (writable copies, only the differences stored) are incredibly useful.
``There are no arbitrary limits in ZFS.
Heh. I guess 64-bit isn't arbitrary anymore?
I wonder about:
``if you enable compression on a swap volume, you now have compressed virtual memory.''
How about encrypted virtual memory?
Finally, I'm curious to see how this will stack up against Reiser4 in terms of features, performance, and everything.
Please correct me if I got my facts wrong.
Man pages and a PDF slide show convering of the more interesting points:
o _ufs_performance_comparison
s trating_zfs_self_healing
s aves_the_day_ta
s _from_a_ras_point
s _that_it
a ge_zfs_from_your_browser
s _the_benchmark
_ boot
http://www.opensolaris.org/os/community/zfs/docs/
(ZFS itself has just two commands btw)
Some basic UFS vs ZFS benchmarks:
http://blogs.sun.com/roller/page/roch?entry=zfs_t
(I guess we'll have to wait and see if ZFS can beat UFS on all benchmarks by the time it ships with Solaris proper)
Party trick - silently recovering forced data corruption:
http://blogs.sun.com/roller/page/timc?entry=demon
A user example of how ZFS's built-in error detection and correction can find hardware errors:
http://blogs.sun.com/roller/page/elowe?entry=zfs_
Some background on RAS in file-systems:
http://blogs.sun.com/roller/page/relling?entry=zf
ZFS vs Veritas for simplicity:
http://blogs.sun.com/roller/page/timf?entry=zfs_i
You can config ZFS from a browser too if you want:
http://blogs.sun.com/roller/page/talley?entry=man
How to trash your OSs with benchmarks:
http://blogs.sun.com/roller/page/bill?entry=zfs_v
Can't yet be used as the boot file-system, but it's being worked on:
http://blogs.sun.com/roller/page/tabriz?entry=zfs
Er, did you read the article the same article that I did? The filesystem apparently supports transactions both for file contents and filesystem structures. They are bragging that they have simulated crashing the computer over a million times without corruption. Sounds like they have it covered.
sheep.horse - does not contain information on sheep or horses.
It's good that Solaris is heating up again. Linux has this "OS for all seasons" thing going on, and needs healthy competitors in all of its areas of utility to keep things real. Competition is GOD in the software and computer hardware world -- even in the OSS world where competition and collaboration run hand in hand.
Dan Price has just put up a new screeen cast demoing trashing one side of the mirror without harming the contents. It's at http://www.opensolaris.org/os/community/zfs/demos/ selfheal
Tp.
Sun releases a staggeringly cool file system, and nobody knows about it.
/. really sucks sometimes.
Only 25 comments too. Apparently there is a definite audience to cater to now rather than providing actual news. Mustn't frighten the linux weenies. *sigh*