Thats just not true............a snapshot (in ZFS) is fully browsable and individual files can be copied out of it as easy as copying the file from one place to another......if you run ZFS cd into the .zfs folder of your pool and have a browse around........Its hugely flexible......nobody has mentioned cloning yet either.......
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Shuts Down Open Source ZFS Project
- Thread starter Small White Car
- Start date
- Sort by reaction score
Thats just not true............a snapshot (in ZFS) is fully browsable and individual files can be copied out of it as easy as copying the file from one place to another......if you run ZFS cd into the .zfs folder of your pool and have a browse around........Its hugely flexible......nobody has mentioned cloning yet either.......
Yes that's kind of my point. Typically with snapshots you have to snap ABCDEFG files when you may really only want to save file "D". I think the area that Apple wants to stick with is accounting for every file and not having to backup more than what is necessary.
At an Enterprise level this would be hard and probably not desired considering the amount of data but for a consumer or workstation system the managing of thousands of files would be easier IMO. Sorry if I didn't clarify myself well enough. I like snapshots overall.
What annoys me is that you cant just decide "oh, we're going to invent a new filesystem today". No matter how smart the people you hire, there is no way Apple is going to turn around and fully implement a brand new (stable) filesystem within the next... what... 18 months for 10.7?
But, maybe we are over-reacting slightly. ZFS is a filesystem and volume manager all-in-one... so maybe Apple wont go down that road and it will be a lot simpler than I imagine.
But, maybe we are over-reacting slightly. ZFS is a filesystem and volume manager all-in-one... so maybe Apple wont go down that road and it will be a lot simpler than I imagine.
But its all block level, when i snaphot a 5Tb filesystem the snapshot consumes only a few Mb...As the main "live" filesystem changes (has files deleted anyway) effectively the snaphost grows but its all block level so not by as much as you would think.....If you save a file 100 times in differnet places it doesn't take up 100* the space if you see what i mean.......If you have a 1Gb file and add 4k to it in a seperate copy of the file you have increased your storage need by 4k ...not 1Gg+4k
Here's why I'm optimistic.
When Apple moved to an OS based on NeXT they had larger quandry than changing filesystems.
They had fs and application problems that existed down to the kernel and frameworks to develop apps.
They came up with Carbon. It deprecated API that would not be able to make the transition to the new OS.
I think the same thing is going to happen with the fs. HFS+ will likely not get replaced by an entirely new beast but it will probably undergo heavy rearchitecture. Apple's worked hard to add features to HFS over the years and a stripped down and rebuilt model may provide what they need.
Most of us here aren't filesystem experts but one thing is sure the new filesystem only need support 10.6 and on because Apple has already taken steps to move away from PPC legacy and onward with Intel.
When Apple moved to an OS based on NeXT they had larger quandry than changing filesystems.
They had fs and application problems that existed down to the kernel and frameworks to develop apps.
They came up with Carbon. It deprecated API that would not be able to make the transition to the new OS.
I think the same thing is going to happen with the fs. HFS+ will likely not get replaced by an entirely new beast but it will probably undergo heavy rearchitecture. Apple's worked hard to add features to HFS over the years and a stripped down and rebuilt model may provide what they need.
Most of us here aren't filesystem experts but one thing is sure the new filesystem only need support 10.6 and on because Apple has already taken steps to move away from PPC legacy and onward with Intel.
Apple does not need a new FS right away. The current file system is fine. So many of you technology fanboys are always looking for the next shiny.
I disagree, the next shiny already exists and it's a landscape changer....ZFS opens up a shedload of niceness.....it would just be nice not to have to deal with solaris (or dodgy looking ports) to get your hands on it. It has it's fans but to me ZFS & Solaris is like buck rodgers riding a knackered old horse.
Much of what Apple has done with OS X is make it hardware agnostic, so they can move in what ever direction they need, and I'm sure all the techies here will claim it make programmers write better code to.
So I wonder if the next step is not a new File system but to sideline the FS from the OS?
To me the clues are.
Core Data.
Core Data manages a pool of data objects in memory looks after writing them to disk, looks after undo/redo(i.e.backup).
So instead of HD why not have a pool of Flash (or one of the faster alternatives set to come on to the market soonish), Store all the data.
C-Blocks.
In cocoa everything is data, with blocks even more so. look at security features anonamising code placement now when a program is loading. A program could be complied to be a group of blocks managed by Core Data Store. As code is called core data pulls the block out of store adds the data to and dumps it into the right queue at Grand Central.
Pull these together with the hardware change bring a big pool of Hybrid memory into a memory bus. Then the HD and FS just aren't needed.
Plus the power use, with zero no time and no HD would be impresive.
Except for say time machine or longer term backup, reading new data, write achieves for storage. All of which could be handled in Usr space like FUSE. No more or less important than a web browser or your email client.
So I wonder if the next step is not a new File system but to sideline the FS from the OS?
To me the clues are.
Core Data.
Core Data manages a pool of data objects in memory looks after writing them to disk, looks after undo/redo(i.e.backup).
So instead of HD why not have a pool of Flash (or one of the faster alternatives set to come on to the market soonish), Store all the data.
C-Blocks.
In cocoa everything is data, with blocks even more so. look at security features anonamising code placement now when a program is loading. A program could be complied to be a group of blocks managed by Core Data Store. As code is called core data pulls the block out of store adds the data to and dumps it into the right queue at Grand Central.
Pull these together with the hardware change bring a big pool of Hybrid memory into a memory bus. Then the HD and FS just aren't needed.
Plus the power use, with zero no time and no HD would be impresive.
Except for say time machine or longer term backup, reading new data, write achieves for storage. All of which could be handled in Usr space like FUSE. No more or less important than a web browser or your email client.
I made a comment earlier but maybe I can help shed some light on why I think a filesystem like HAMMER(http://en.wikipedia.org/wiki/HAMMER) is a good fit for apple.
This all comes down to licensing issues. Btrfs looks awesome and many of the features dovetail those of ZFS but Btrfs is really for linux and not for a BSD like operating system. In fact, if you look at what filesystem are supported by FreeBSD you would be pretty underwhelmed (UFS, ZFS, HAMMER). Linux has tons of filesystems but I believe most of them are GPL'd and wouldn't work for apple.
Apple wants to get rid of their core systems running on GPL code. For poof of that just look at LLVM. They are trying to phase out the GCC compiler in favor of one that is based on the BSD license. That means that their kernel, compiler and file system will all be BSD licensed and they don't have to worry about any GPL issues.
Now, I have no idea if HAMMER will actually be used by apple. I know DragonflyBSD (a fork of FreeBSD) is using it and it's still under pretty heavy development. But I wouldn't be surprised if apple forked it, had it's engineers add a bunch of code and then release it back to the community and also piss off the original devs. The same thing happened with KHTML/WebKit.
This all comes down to licensing issues. Btrfs looks awesome and many of the features dovetail those of ZFS but Btrfs is really for linux and not for a BSD like operating system. In fact, if you look at what filesystem are supported by FreeBSD you would be pretty underwhelmed (UFS, ZFS, HAMMER). Linux has tons of filesystems but I believe most of them are GPL'd and wouldn't work for apple.
Apple wants to get rid of their core systems running on GPL code. For poof of that just look at LLVM. They are trying to phase out the GCC compiler in favor of one that is based on the BSD license. That means that their kernel, compiler and file system will all be BSD licensed and they don't have to worry about any GPL issues.
Now, I have no idea if HAMMER will actually be used by apple. I know DragonflyBSD (a fork of FreeBSD) is using it and it's still under pretty heavy development. But I wouldn't be surprised if apple forked it, had it's engineers add a bunch of code and then release it back to the community and also piss off the original devs. The same thing happened with KHTML/WebKit.
I'm sorry but you are just a fanboy of new and shiny technology just of the sake of newness. I've taken a look at ZFS and not only does it not make sense for the desktop paradigm but it sounds user unfriendly and unintuitive.
Does a user really want to have wrap their head around the concept of a "pool"? Would a regular user understand that a pool would get broken if you removed a USB stick in pool?
ZFS sounds nice for a Storage Area Network or SAN for short but it really has no business being on a desktop computer for regular users.
HFS+ has all of the features the current Snow Leopard OS needs.
Have you forgotten who runs Oracle? Its former Apple board member and Jobs best friend Larry Ellison. If Jobs and Apple wanted ZFS, they could get it.
It sounds like Apple knows something that the public doesnt know yet.
Isn't the real reason obvious to everyone?
The Iphone OS doesn't need ZFS.
End. Stop.
The Apple Consumer Gadget Company doesn't need ZFS on MP3 players. Gone. End of story.
What was the "high end" computer announcement from Apple this week?
- Yes, now the mini has two laptop drives and they call it a server.
Apple is focused on people who want to buy shiny things on the Internet. If you want professional hardware and software - the writing is on the wall, start looking around for a new vendor.
There's a small problem here - Ellison can't do diddly, Oracle does not own Sun.
The Iphone OS doesn't need ZFS.
End. Stop.
The Apple Consumer Gadget Company doesn't need ZFS on MP3 players. Gone. End of story.
What was the "high end" computer announcement from Apple this week?
- Yes, now the mini has two laptop drives and they call it a server.
Apple is focused on people who want to buy shiny things on the Internet. If you want professional hardware and software - the writing is on the wall, start looking around for a new vendor.
There's a small problem here - Ellison can't do diddly, Oracle does not own Sun.
ZFS would have been a low net effort solution without the time sensitive endpoint. Oracle is adverse to Apple so even if Sun had a tentative agreement with Apple any such contract has bail provisions Oracle would have found nitpicky ways to invoke. Once Oracle bought Sun, ZFS as great as it is, became a big bag of hurt. Politics and personalities trumps technology.
I said right here a couple of years ago that Apple most certainly should have bought Sun. Sun would have approved it too.
Then Sun division of Apple could have built out all that AT&T wireless stuff they needed.
Rocketman
Woulda shoulda.
I said right here a couple of years ago that Apple most certainly should have bought Sun. Sun would have approved it too.
Then Sun division of Apple could have built out all that AT&T wireless stuff they needed.
Rocketman
Woulda shoulda.
Well, you can look forward to Microsoft's 128-bit filesystem....
I bet microsoft will just 128bitify NTFS. I mean, theres not a lot wrong with NTFS, it just has trouble with fragmentation and a few other nuances. ZFS 'looked like' a fusion of the best of NTFS and the best o UNIX like FSes.
Maybe Apple will buy source for XFS or whatever the next gen Linux one was?
And as server doesnt have to be a large bulky thing Aiden. I mean, my sisters file server is a Shuttle with two 1TB drives in it.
Yea, it has everything except built-in snapshots, copy-on-write, and full checksumming of every block. Sounds like a great filesystem to archive data on! I'm an OS X "home user" and all of my data sits on my OpenSolaris server with ZFS. No way in hell I'm trusting HFS+ to archive data, especially stuff I buy and officially cannot download again.
Why on earth should Apple have bought Sun? For the massive quarterly losses? Just for a better license on ZFS? Ridiculous.
As long as Microsoft fixes that nonsense that a fully qualified filename can only be 32K characters long. And the ultra-nonsense that some older APIs limit the filename to 260 characters.
What a hold-over from the floppy disk days....
Oh - alert for Dmann/LagunaSol - Aiden has made a post critical of Windows. What do you do now?