Back in May of this year it was rumoured that Apple’s File System Development Team had contacted Sun Microsystems to help in a translation of the Zettabyte File System to Mac OS X.
With the most recent Build of Mac OS X 10.5 Leopard, ZFS has appeared to have finally made an appearance, as per the rumours.
Sun Microsystems describes ZFS as the “last word in File Systems,” the file system is well known for being incredibly resilient to data failure, having immense capacity and being the first 128-bit file system.
Lets look at ZFS’s ability to keep data safe, everything on ZFS is “copy-on-write” meaning that the file system never overwrites live data and it leaves no windows of vulnerability. Also on ZFS, everything is transactional meaning that a file copy will either completely fail or be a complete success not half and half leaving you with a mess to clean up.
Finally the most important feature of data integrity os check summing and everything on ZFS is checksummed meaning zero data corruption.
The Zettabyte File system is also incredibly scalable adding value to users of both server and home environments. ZFS has a data storage capacity of 256 quadrillion zettabytes and a maximum file size of 16 exabytes (that’s 17179869184 terabytes).
Overall ZFS offers the following key advantages:-
- Pooled storage - No requirement for a volume manager when extra volumes added, the volume is simply added to a pool creating a vdev (virtual device), a collection of vdevs makes up a zpool, which in essence is the storage available to the file system.
- Snapshots - Read-only point in time of the file system
- Clones - write-able copy of a snapshot
- RAID-Z - Makes use of copy-on-write; rather than overwriting old data with new data, it writes new data to a new location and then overwrites the pointer to the old data
- Detects and then corrects data corruption
- Incredibly fast due to intelligent pre-fetching, and dynamic striping.
You can view other Leopard screenshots from previous World of Apple exclusives, here and here.
Comments and Trackbacks
All comments made are owned by their authors. Please keep discussion clean and relevant to the main article. Basic HTML tags can be used for formatting comments, and avatars are provided by the Gravatar service.
Trackback link for this entry | RSS Feed for comments
The following sites have trackbacked to this entry:
The following comments have been added by readers:
Luke Lonergan
18th December 2006, 06.23 am
Hooray for Sun/ZFS and Apple! ZFS will make an outstanding desktop filesystem. One of it’s built-in management features allows you to “snapshot” points in time as backups, which sounds very much like apple’s “time travel” feature - I wonder if there’s been some discussion there?
- Luke
Quote | Comment
Leon Koll
18th December 2006, 08.47 am
“This post is open for both pinging and commenting.”
I know how to comment, but don’t know how to ping the post. Could you please elaBorate…
Quote | Comment
sonofshatner
18th December 2006, 14.39 pm
Ok - it is always cool when a ‘better’ technology gets included, but what I can’t stop asking myself is this… Why did they do it? Apple usually only makes moves like this when there is serious product around the decision. So my other question is when will this product be avail? Is it a XServe Raid ZFS based on Leopard…?
if you looking for a ps3 or wii - i have helped a few dozen folks find em, at my site.
Quote | Comment
Alex Brooks
18th December 2006, 14.43 pm
Leon,
To ping a post you usually link to it via whatever blogging platform you are using, this will ping the post and appear here as a comment.
Quote | Comment
t3knomanser
18th December 2006, 15.04 pm
Hoo rah. That’s frickin’ great. Guess I won’t need to take that old desktop to set up a Solaris file server at home any more. I’ll just get me a Mac Mini.
Quote | Comment
aboyko
18th December 2006, 15.07 pm
Hope they’re sticking close to the current code; lately ZFS has added features like hot spares and RAIDZ2 (~RAID-6, two parity disks). Seems to keep getting better and better. (I also had a plan to convert a home storage server from Linux to Solaris Express; now I can just keep it all-Mac…)
Quote | Comment
Patrick Giagnocavo
18th December 2006, 16.38 pm
As a user of Solaris on many customer sites, let me say that working with ZFS is great. Excellent performance as well.
Quote | Comment
Tim Ellis
18th December 2006, 16.52 pm
Interesting, can it boot from a ZFS partition?
Quote | Comment
Wes W
18th December 2006, 18.40 pm
Solaris can’t yet [officially] boot from ZFS, so the odds that OS X is booting from ZFS is slim to none. However, ZFS is an awesome filesystem and anyone with an OS X non-root storage array will probably want to convert. I’ve gone ahead and built a ZFS stoarge server running OpenSolaris for my OS X machine, though native support for ZFS on the Mac is great news, though this brings up another question…When will Apple finally take all of Solaris and build it’s GUI easy-to-use-goodness on top?Now that would be an OS to take over in the corporate environments as well as at home.
Quote | Comment
Gerrit DeWitt
18th December 2006, 19.05 pm
*ZFS has a data storage capacity of 256 quadrillion zettabytes and a maximum file size of 16 exabytes (that’s 17179869184 terabytes).*
Practically speaking, ZFS has a maximum volume and maximum file size of 16 exabytes (EB), and that’s actually not equal to 17179869184 terabytes:
In decimal, one exabyte is 10^18 bytes (one quintillion bytes, using US English place value names), and one terabyte is 10^12 byes (one trillion bytes, US English place values). In terms of usable space (capacity), one binary exabyte (exbibyte, EiB) is 1024^6 = (2^10)^6 = 2^60 bytes, and one binary terabyte (tebibyte, TiB) is 1024^4 = 2^40 bytes.
A little stochiometry tells us:
1. 16EB * (10^18 bytes/EB) * (10^-12 TB/byte) = 16 * 10^6 TB = 16,000,000 (sixteen million terabytes in 16 exabytes).
2. 16 EiB * 2^60 bytes/EiB * 2^-40 TiB/byte = 16 * 2^20 = 2^4 * 2^20 = 2^24 = 16,777,216 (sixteen million seven hundred seventy-seven thousand two hundred sixteen binary terabytes in 16 binary exabytes).
3. 17,179,869,184 TiB is actually 16 binary zettabytes (2^70 bytes). One zettabyte is roughly one sextillion (US English units) bytes, 10^21. Even though the filesystem’s name is ZFS, the limitations are presently in exabytes (see the Wikipedia article):
17,179,869,184 TiB * 2^40 bytes/TiB * 2^-60 EiB/byte = 16,384 EiB
17,179,869,184 TiB * 2^40 bytes/TiB * 2^-70 ZiB/byte = 16 ZiB
More info:
http://en.wikipedia.org/wiki/Zfs
Quote | Comment
Alex Brooks
18th December 2006, 19.30 pm
Wonderful insight, Gerrit.
Would you mind taking a look at the following PDF (http://www.opensolaris.org/os/community/zfs/docs/zfs_last.pdf), page 19 lists the figures I gave, how do they match up with those of Wikipedia?
Quote | Comment
spatialguru
18th December 2006, 22.33 pm
So that’s how they will handle the underlying requirements of their time machine technology.. I was wondering how they’d make it rock…
http://www.apple.com/macosx/leopard/timemachine.html
Quote | Comment
MacFSGuru
18th December 2006, 23.19 pm
To spatialguru: Time Machine is built solely on HFS at this point. It is likely that ZFS will eventually be supported.
Quote | Comment
vertgo
19th December 2006, 05.01 am
Finally! Next generation self healing raid storage here I come!
Quote | Comment
osgeek
19th December 2006, 21.14 pm
Good info on ZFS. I like the file system myself.
Quote | Comment
Springfieldian
20th December 2006, 20.52 pm
I think it looks XSANtastic.
Quote | Comment
Fred
21st December 2006, 08.33 am
Looks like a fake. I’m using Leopard Build 321 on PPC, and there’s no ZFS support there in Disk Utility. In fact, in that build they’ve removed UFS and FAT32 from the options, so there’s only HFS with and without Journaling.
Unless this is from the latest Apple Internal build (326), THIS IS FAKE.
Quote | Comment
duffman
22nd December 2006, 14.16 pm
“Unless this is from the latest Apple Internal build (326), THIS IS FAKE.”
Or its intel only …
Quote | Comment
Alex Brooks
24th December 2006, 12.22 pm
Fred,
This feature is being touted by many rumours sites around the web, many of them showing different methods of getting to the feature.
I can’t possibly see how you’ve come to the conclusion that this is fake.
And for the record, build 326 never existed, not even as an internal build.
Quote | Comment