“If there is a defining feature in Apple’s Leopard, it’s Time Machine. As cool as it may be, the fact is that we, as Linux users, are obviously not going to see much benefit from this is pitiful. So it’s a good thing that open source developers have taken it upon themselves to create something similar, be it not a ‘pretty’ alternative.”
wow, that review/article/site was terrible. not only was it plastered with ads, but it was so small and didn’t actually provide me any meaningful details.
What ads?
Indeed, I guess konqi’s addblock feature works as advertised (pun intended)…
We have lots of options in the open source world. The only difference is, there isn’t an ultra pretty and snazzy GUI for what we have available. You simply aren’t going to get that kind of 3D Time Machine GUI alternative using GTK.
If you put a reasonable GUI onto Bacula then you’d have something the average person could use, and you’d have rather a lot of enterprise backup companies rather worried. If you combine that with LVM snapshots that can enable live backups then you’ve got it made really. Anything else is going to be prone to consistency problems.
Do grow up from linux playground and try to understand something about solutions.
Time Machine is not a gui!
It is an idea and easy solution: backup. Create snapshots and backup. Something that works, easy and no bs with setup. Something that just works, something that my mother can set up, not only restore from.
The only one that can pull this out in linux world is Ubuntu. Sorry, redhad/suse might also pull it off, but that is improbable.
Hard linked copy (current time machine backend) is a quick hack instead of zfs, zfs is quite good and Sun is bragging for a very good reason. I have used it on Solaris, it is good. It is probably the reason why you should make that next 2TB home file server on Solaris.
I bet Time Machines idea came from zfs, zfs has the idea of unlimited snapshots.
Hard linked copy (current time machine backend) is a quick hack instead of zfs, zfs is quite good and Sun is bragging for a very good reason. I have used it on Solaris, it is good. It is probably the reason why you should make that next 2TB home file server on Solaris.
I bet Time Machines idea came from zfs, zfs has the idea of unlimited snapshots.
Snapshots might not really be that useful on a home file server, but zfs would have lots of other useful features there. I especially like that you can just add a new hdd, add it in the spool and it’s instantaneously useable. No need for formatting, resizing partitions or any such hassle. And it’s self-healing too. I just don’t know if my home server would be able to run Solaris.. It’s just a 1ghz Athlon with 128mb RAM :/ (and I have absolutely no experience whatsoever with Solaris)
While it won’t work with just 128 MB RAM (1 GB is the recommended minimum), you can use ZFS with FreeBSD 7. Solaris is not the only option.
Regardless of which OS you run it on, the best setup for ZFS is to use a 64-bit CPU and a 64-bit OS, with 2 GB or more of RAM.
While it won’t work with just 128 MB RAM (1 GB is the recommended minimum), you can use ZFS with FreeBSD 7. Solaris is not the only option.
Regardless of which OS you run it on, the best setup for ZFS is to use a 64-bit CPU and a 64-bit OS, with 2 GB or more of RAM.
Thanks, but that’s just exactly as I thought
But hmm, IMHO such a machine is a bit overkill for home file server :3
Exactly. I have read a lot about ZFS and got pretty excited about it since I have been a victim of bit rot.
I rotate my machines out. My best machine is used as my main desktop machine, usually dual boot between Linux and Windows (for games). My second best machine is used as my file server, and the others just sit there and I use them for spare parts diagnosing problems with my friend’s computers.
The requirements for ZFS are nuts, my fastest machine right now is only a P4 3GHz HT, not 64 bit and only has 1Gb of memory. I don’t need super high performance from my fileserver so I could probably try it but this machine is also used for MythTV. If I could run ZFS under Linux I’d give it a try but I don’t think I can run MythTV under Solaris or FreeBSD.
Do grow up from linux playground and try to understand something about solutions.
Grow up Mac boy and understand why Time Machine is even remotely useful to people.
Time Machine is not a gui!
The only useful part of Time Machine is the GUI, which makes it readily usable in a way that other backup solutions are not. Seriously, the backend infrastructure to Time Machine is two a penny in the open source world, and quite frankly, Time Machine isn’t even all that good. Hard links are bloody awful, and they will go wrong.
Create snapshots and backup. Something that works, easy and no bs with setup. Something that just works, something that my mother can set up, not only restore from.
Because Time Machine is pre-installed and has a GUI that makes that possible in a manner that users can see and understand. They’re going back in time……
The only one that can pull this out in linux world is Ubuntu.
Why? I’m greatly amused by this strange idea people have that Ubuntu will be able to achieve all this. What was described in that article is by no means spectacular or remotely as decent to use as the Time Machine GUI, and is a very, very long way off.
zfs is quite good and Sun is bragging for a very good reason. I have used it on Solaris, it is good.
ZFS has snapshots, and in the Linux world we have LVM snapshots. The article was talking about how we could have something like Time Machine in the Linux world. In reality, we could quite conceivably have something far better with a front-end to match. It won’t be achieved by doing what they’re doing.
Time Machine is not a gui!
Yes it is to a large degree.
Currently it is a 3D Gui + as you said a Hard linked copy (current time machine backend)
From the Flyback website:
http://code.google.com/p/flyback/
Apple’s Time Machine is a great feature in their OS, and Linux has almost all of the required technology already built in to recreate it. This is a simple GUI to make it easy to use.
In the sites Wiki
* yes, an 3D/opengl view of the directories “flying back” is coming… =P
So it will soon have the glitz of the OSX version.
This reminds me of the whole Exposé and Komposé fracas. Panther came with Exposé, and it took KDE a few months to knock something up that looked vaguely similar. Then despite the fact that the performance was utterly lacking and that it didn’t windows didn’t do live updates, Komposé was hailed as an alternative to Exposé. So we had the ridiculous threads where people would keep claiming that Komposé was equal in functionality to Exposé, all the while ignoring that Exposé works fine on weak machines (used it lots on a 867Mhz G4 powerbook) and it did live updates.
All these Time Machine “alternatives” just gave me a deja vu moment.
It’s fun to see all these software trying to mimic the Time Machine smoothness and ease of use. But they are all lacking one little thing.
Leopard changed a little something to the filesystem allowing to create hard links to directories.
Other incremental backup systems like rsync are able to create hard links to file. So the user is able to browse its entire backup while only taking as much space as what actually changed since the last backup.
But this hard linking does not work for directories in rsync, so each new incremental backup must recreate the whole directory structure (around 600 000 directories on my current Leopard system).
These directories take a very large amount of drive space, compared to what actually changed in the data, so backuping with these tools is not done frequently in order not to waste to much space with these directories. Beside, they often do not backup the whole hard drive but only “important” data.
Time Machine is able to create hard links for directories where no change were made, just creating one link instead of a whole directory structure. It allows Time Machine to quickly backup your entire hard drive every hour, without taking too much space on your backup drive.
Until an implementation of Time Machine supports directory hard-linking, I doubt they will be able to offer the same experience as Time Machine does.
What you are talking about is pretty much just a hack to get around not having support for snapshots in the file system.
NTFS and ZFS for example both support snapshots. Time Machine probably will too when MacOS switches to a newer file system like ZFS.
It will be interesting whether Apple’s storage products, if any are released at MacWorld, which will take advantage of ZFS. It would be a great win for those who are looking at an alternative to the big players out there.
NTFS and ZFS for example both support snapshots. Time Machine probably will too when MacOS switches to a newer file system like ZFS.
FreeBSD UFS filesystem got snapshots for ages- I guess same technology is used in OS X.
http://people.freebsd.org/~rse/snapshot/
FreeBSD UFS filesystem got snapshots for ages- I guess same technology is used in OS X.
Naah! it’s the same technology as used in Linux both AFAIK use rsnaphot a PERL script front end to rsync to acheive this.
Flyback is a GTK-Python program for this based on rsync.
Tridge rules!
Naah! it’s the same technology as used in Linux both AFAIK use rsnaphot a PERL script front end to rsync to acheive this.
Naah! – I was probably wrong. It looks like your reference doesn’t use rsync.
However rsnaphot as a front end script to rsync in order to create snaphots has been around for BSD and all unixlike operating systems for some time.
Time Machine seems to be an idea grown out from zfs.
Please try to see the limitations of time machine because the backend is not zfs. If it would, then, well, the whole Time Machine would be very different! Something that neither Linux or Win can achieve. You would always use all your local hdd for previous revisions, not only external hdd. You would have always 10min versions of all the files up as much free space you have.
This is what I had hoped from Leopard.
How does Apple claim OS X is a unix and managed to have two folders hard-linked with the same inode number?
As far as I know, no other Unix does this and it seems to me that this might lead to some serious issues with fire manipulation tools that do not expect this type of behavior.
Well obviously they manage somehow since the Leopard installations out there haven’t all blown up in the meantime, don’t they? Unix shmunix, as long as it works.
Not a lot of applications work with inodes directly anyway, do they?
Good point, I recall there are serious issues what that feature – which is why it isn’t in any other modern unix that I know off. It even had to be pulled out from Reiser4 to give it a chance of entering the Linux kernel. So either the apple engineers did fix the issue in a brilliant way (so we should look at it and steal it) or they just ignored it, and try to work around the issue… I wonder 😉
There are lots of options for doing backups in linux. One of the best is Mondo Rescue. It does incremental and system backups and restores. Of course, since Linux supports LVM, you can also do full backups while you are working, Of course, you can also set up raid one fairly easily if you just need redundancy.
Yes, rdiff-backup is a very good choice. As opposed to some of the solutions mentioned in the article it handles ACLs, SELinux, and other extended file attributes.
Another idea to make your life easier would be to create nautilus action “Replace with latest version from backup” so that you easily can get back to the latest backed up version in case you get some unrecoverable error in your document. This would of course work with almost any type of backup system that is scriptable.
If you want something more advanced you could let the nautilus action be a tcl/tk script that provides a GUI to let you choose a date. If I remember correctly there is also some gnome tool that alows you to create simple gnome GUIs from a shell script, but the name have slipped my mind right now.
Another idea for easy backups would be to let incrond monitor changes in e.g your home directory and have it initiate some backup program, simply just make a copy of the file to a backup location it is changed.
I wonder how close Subversion is to being able to snapshot an entire directory, complete with device nodes, pipes etc.
I have been using svn for ~/ since late 2004. Very easy to set up and keep synced for ~/doc, ~/tunes, ~/src, ~/lib, ~/media, etc. Makes it rather easy to do updates and keep the laptop and desktop on the same page.
Is there a way to configure svn to do not a local copy to .svn ?
If not, svn would use too much disk space, I am realy afraid.
I do backups using rdiff-backup… This program allows you to easily maintain a mirror of your last backup, plus a compressed, diff-based history of all previous backups. This makes it possible to take very frequent backups, space efficiently, and to pull tricks like “get me this file, as it was on this date”.
To me, rdiff-backup seems like the best backup system for a desktop user that I’ve seen so far. What it is, of course, missing is a well-established friendly GUI. The KDE program “Keep” does a decent job of it though, giving friendly access to the powerful backend.
Think Similar (TM)
to do incremental backups aka snapshot. One ZFS blogger took snapshots each 10 minute or so, resulting in ~6.000 snapshots in some period of time. It turned out that there was only ~4GB of data that had changed in those 6.000 snapshots.
ZFS makes it very natural to do incremental snapshots, without extra space.
Edited 2008-01-06 11:39
ZFS indeed is extremely handy, taking snapshots is so extremely easy, you can do full backups too if you wish, you can even work on the files inside backups and snapshots just as you would on any other file. Speak about a very powerful backup system indeed!
I just so much wish it was useable under Linux… 
LVM on Linux also supports snapshots. The can be created instantly, and only take as much space as the number of changed blocks (copy-on-write). Snapshots can be read-only or read-write (useful to save space in a virtualization/containers/openvz environment).
The only limitation is that you can’t take snapshots of snapshots, but that may change in the future.
LVM on Linux also supports snapshots
Yeah, but you need to also reserve space where to save those snapshots. If you’ve got a partition which spans the whole LVM volume you just don’t have anywhere to save snapshots to. That’s the power of ZFS spools: all the storage space is available at all times and only consumed when actually needed.
Nice to see Linux developers hard at work copying yet another feature from other OSes :-). No wonder they’re being hit with patent lawsuits up the wazoo.
In fact, Linux never has been hit with a patent lawsuit up to date (January 2008).
And innovation is poured into Linux all the time. Think Reiser3, XFS, schedulers, …
Of course as any good product Linux incorporates known good ideas as it incorporates completely new features. And the fact, that lots of these good ideas are brought to Linux by corporations who do not recieve monetary payment should show everybody that the open source method is the best available method for developing a kernel. SGI even gave up their own kernel because participating in the Linux development was by far cheaper than maintaining their own kernel development.
I would like to have all good features that someone can think of in Linux, be they completely new (like the CFS scheduler) or be they inspired by something else.
If the world actually worked this way, neither MS Windows OR MacOS would ever have made it through the first year. Apple and MS would have been sued into oblivion by Xerox. Gary Kildall would have had a piece of MS too over the fact MS-DOS was nothing but a lame CP/M clone.
We’d all be using Xerox workstations. Remember the Xerox Alto?
Software patents are a recent evil. One that is fortunately doomed to fail. It has already most everywhere except here. Keep your trolling to yourself. BTW, there hasn’t ever been a SUCCESSFUL lawsuit against Linux. Some trolls have tried. All have failed.
–Kevin
The best open-source backup software is backup-manager in my opinion. I use it on all my servers and desktops: http://www2.backup-manager.org/
Until Linux has a working ZFS implementation (which would have to be under FUSE), this Time Machine stuff is useless.
It is useless under OSX too for now, but I’m sure the ZFS implementation is in the works and won’t have to run in userspace.