Time Machine is Missing a Trick
The first thing I’ve played with since updating to OS X Leopard is Time Machine and I have to admit I’m slightly disappointed.
Time Machine is a fantastic addition to any operating system, and it’s brilliant and very cool how it makes backing up and restoring so easy, however it’s using a standard Journaled HFS filesystem (from the output of mount):
/dev/disk2s3 on /Volumes/TimeMachine (hfs, local, journaled)
(I called my backup disk TimeMachine for easy identification).
Why isn’t Time Machine using ZFS? Apple advertised they would include ZFS support in OS X, but only seem to have offer read-only support, and haven’t actually taken advantage of this brilliant filesystem in anyway.
A bit of searching seems to indicate one of the reasons they can’t use ZFS just yet is because Time Machine takes advantage of directory hard linking, which ZFS doesn’t support (at this time).
I know disks are cheap these days, but using hard links seems a bit wasteful. What if you have a 2Gb file and all you change is 10 bytes in the file? The WHOLE file is backed up again. Surely using something like the ZFS snapshot functionality would have been a better idea.
I would have thought the snapshot functionality of ZFS would have been ideal for Time Machine, but I guess Apple have other ideas. Who knows, maybe they’ll change this when read/write ZFS support is officially supplied.
Talking of ZFS, time for me to download the ZFS Beta Seed and bring all the joys of fully functional ZFS to OS X.
3 Responses to “Time Machine is Missing a Trick”
Speak Your Mind 
Search
Related Entries
- Data Centre In A Box
- 64-bit Flash Player For Linux - Why?
- On the Road to Ruby on Rails
- OpenSolaris 2008.11 is Now Available
- New 24″ iMac on Order
- Links for 6 Apr 2009 - 3 May 2009
- HOWTO: Disable Timeout Alert in Lightning/WCAP Extension
- Thumper From Sun is A Beast
- Links for 13 Jan 2008 - 27 Jan 2008
- RedHat Kernel Crash Dumps


I agree, it could be better.
I have also been a bit disappointed by the way Time Machine is working. As part of my
blowing up my system using ZFS series I tried a restore.
For some reason and without much in the way of feedback there was a problem either
reading or writing a file. This aborted the whole restore 8GB into a 48GB restore.
I’ve moved to rsync for the restore to see if there are still “problems”
As for the links, have you looked at the extended attributes?
drwxr-xr-x@ 6 root sferry 204 Nov 30 01:30 2007-11-30-013037
com.apple.backup.SnapshotNumber 3
com.apple.backupd.SnapshotCompletionDate 17
com.apple.backupd.SnapshotStartDate 17
com.apple.backupd.SnapshotState 2
com.apple.backupd.SnapshotType 2
./iTunes Music/Various/Dr. Demento Presents The Great:
total 7752
-rw-r–r–@ 1 sferry staff 3965305 Dec 25 2006 05 Gayla Peevey - I Want a Hippop.mp3
com.apple.metadata:_kTimeMachineNewestSnapshot 50
com.apple.metadata:_kTimeMachineOldestSnapshot 50
I assume that the metadata is what they use to keep a fully consistent image across the backups
while saving space with the hard links. If you look at that data and think about user properties
in ZFS I think I see an overlap.
Well look at that. I’ll have to look into this in more depth when I get home. I think I’ve got to try my first restore too as since I installed the ZFS series, everytime I power off my external USB hub, OS X crashes :-/
Are you looking at the zpool status before you turn it off?
I found that I could reliably cause a crash every time I disconnected a pool that didn’t have a valid local replica.
e.g. when as I was sloshing data to the external in preparation to establishing a mirror, trying to not require a restore, I accidentally disconnected my firewire cable and crashed.
Similar things playing with the usb thumb drive. If the internal was bad and I disconnected the external it crashed. However if the internal was good and I disconnected the external it was fine.