• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

ZFS To Become Default File System In Leopard

Originally posted by: n0cmonkey
Announcing things like that is the best way to get Apple not to do them. Just ask ATI. 😛
Click to expand...

Yeah that's a pretty major slip-up for an Apple announcement. Bet that guy's head will be rolling when Jobs gives him a call later today...lol.
 
He's high enough up in Sun that he probably won't get canned over it. But if it was true before, it may not be true now.
 
Originally posted by: n0cmonkey
He's high enough up in Sun that he probably won't get canned over it. But if it was true before, it may not be true now.
Click to expand...

I doubt they will ditch ZFS just because someone announced it early, especially with WWDC just around the corner. For something else maybe, but a file system is a pretty integral part of an operating system to give up on in a fit of anger. Plus the rumor sites have been saying ZFS is coming with Leopard for awhile. Besides, it's not like there's anything else better out there - ZFS is killer! 😀
 
It depends on how well ZFS is able to handle the resource forks feature in the current HFS+ file system.

HFS+ is NOT posix compliant. So to support server stuff OS X has always supported UFS. UFS does not support the MacOS resource forks, so it causes occasional application breakage if you use it.

HFS+ has always been a very crappy file system besides that though. It's slow, it's fragile, it's FAT-32 era technology. It supports journalling through the FreeBSD VFS stuff, which is hardly a performance win.

So anything is a improvement. But there is a pretty good chance that Apple will continue to use HFS+ by default and replace UFS with ZFS for server stuff.

 
Originally posted by: Kaido
Originally posted by: n0cmonkey
He's high enough up in Sun that he probably won't get canned over it. But if it was true before, it may not be true now.
Click to expand...

I doubt they will ditch ZFS just because someone announced it early, especially with WWDC just around the corner. For something else maybe, but a file system is a pretty integral part of an operating system to give up on in a fit of anger. Plus the rumor sites have been saying ZFS is coming with Leopard for awhile. Besides, it's not like there's anything else better out there - ZFS is killer! 😀
Click to expand...

No, they'll just move it to non-default.
 
Originally posted by: SoundTheSurrender
Well at least they delivered. Unlike some other company that comes to mind.
Click to expand...

Vista did deliver to many people who bothered to take a closer look. It did not to blind idiots who feed off of negative propaganda and exaggerated problems.
 
I was talking about WinFS.

And no it did not deliver for me. Poor laptop performance, hard drive thrashing at the startup screen, sleep not working, unzipping small files take over 3 minutes.

 
Well at least they delivered. Unlike some other company that comes to mind.
Click to expand...

I guess you're ignoring the fact that ZFS and WinFS had completely different sets of goals. And I guess delivering is an appropriate word for ZFS on OS X because all they're doing is delivering someone else's code with virtually no development done on their end.
 
Originally posted by: mercanucaribe
So??
Click to expand...

What do you mean "So?"

So little contribute? So little understanding? So what?

What?


For mac this is very good news if your a mac user.

ZFS combines raid features, logical volume management, compression, checksums, snapshots, file system features, etc etc into one unified system.

Most of these features can be had in Windows Server. Most of these things can be had in Linux, but in both cases they are quite more complex in terms of code, execution, and management... IF you need these features.

HFS+ and UFS can be considured 2nd generation file systems. Not very sophisticated. These are the defaults for OS X currently.

NTFS and Linux Ext3 can probably be then considured 3rd generation file systems. Current mainstream stuff.

ZFS is next generation, 4th generation. If OS X adopts it and it over comes certain limitations I talked about above then it will be more capable then anything from Microsoft or Linux.

But ZFS isn't established yet. It's not out of beta. It's not tested in the real world by large numbers of people, and it's not going to be paticularly optimized at this point.
 
ZFS combines raid features, logical volume management, compression, checksums, snapshots, file system features, etc etc into one unified system.
Click to expand...

Which is retarded, they're logically separate functions so they should be separate.
 
Originally posted by: Nothinman
ZFS combines raid features, logical volume management, compression, checksums, snapshots, file system features, etc etc into one unified system.
Click to expand...
Which is retarded, they're logically separate functions so they should be separate.
Click to expand...
So... because some people in the Linux community think that these functions should not be integrated, no one using any OS should be able to use something like ZFS?
 
Originally posted by: SoundTheSurrender
I was talking about WinFS.

And no it did not deliver for me. Poor laptop performance, hard drive thrashing at the startup screen, sleep not working, unzipping small files take over 3 minutes.
Click to expand...

What do you mean they 'delivered?' I'm sitting here with a Macbook that's only about 4 months old, and i don't have ZFS on it. In fact, i've been using OSX for several years now, and i don't have ZFS. Exactly when was this 'delivered'?

Of course they're going to upgrade the filesystem... the filesystem they have in now is hardly better than FAT32... and at least Windows has replaced FAT32 many years ago.
 
So... because some people in the Linux community think that these functions should not be integrated, no one using any OS should be able to use something like ZFS?
Click to expand...

No, because unix works so well because it's built from lots of small tools that do one specific job really well. Huge monolithic beasts are much more difficult to maintain, extend, fix, use, etc.
 
Originally posted by: SoundTheSurrender
I was talking about WinFS.

And no it did not deliver for me. Poor laptop performance, hard drive thrashing at the startup screen, sleep not working, unzipping small files take over 3 minutes.
Click to expand...

ROFL. Unzipping small files taking 3 minutes?

Dude you've got a problem there. Why don't you go TROUBLESHOOT it down to ROOT CAUSE before you make some wild assed speculation and try to pass it off as fact that there is a code problem on an OS that's running on millions of PCs all over the world. Give us a break :roll:

If you can't even get NTFS to perform right I would hate to see what trouble you could cause with WinFS.

 
Originally posted by: Nothinman
So... because some people in the Linux community think that these functions should not be integrated, no one using any OS should be able to use something like ZFS?
Click to expand...

No, because unix works so well because it's built from lots of small tools that do one specific job really well. Huge monolithic beasts are much more difficult to maintain, extend, fix, use, etc.
Click to expand...


Yes exactly.

There are a few things lacking with Linux vs Solaris ZFS.

Stuff like online compression and running checksums as well as convient resizing of filesystems and snapshotting. The tools to manage these things are more complex in Linux as far as the end user is concerned. Different utilities such as mdadm vs vgcreate and such things all have different syntaxes and conventions.

This is not a unfixable problem and it just a side effect of this stuff evolving over the years with no unified direction... which is fine because nobody ever realy knows were to go, people have good ideas

Also the time it takes to fsck a file system is excesive and is a problem in enterprise environments for Linux. A multi-terrabyte file system can take days to properly recover.

XFS doesn't have this problem, generally, but it's so complicated that nobody realy knows how it works, at least not in Linux-land. SGI-land, maybe.

But none of this is unfixable.

Once it gets fixed then they will end up being much more capable then ZFS due to the layered design. ZFS will have a much more difficult time coping with changes and new technology as everything evolves due to it's monolythic design. Or something like that.

Sun maintains that while the Unix model of layers is all fine and dandy they designed ZFS by refactoring the basic concepts of software raid and logical volume management with file systems. That the layered approach in this context is a aberration of evolution of storage rather then any sort of logical or correct division. That having it all in the file system context is actually a more correct approach and they actually belong together.
 
Stuff like online compression and running checksums as well as convient resizing of filesystems and snapshotting. The tools to manage these things are more complex in Linux as far as the end user is concerned. Different utilities such as mdadm vs vgcreate and such things all have different syntaxes and conventions.
Click to expand...

The management tool's complexity is orthogonal to how the underlying setup is layered. Nothing is stopping someone from creating an all-in-one RAID/LVM tool to handle all of that. Checksums would probably have to be done inside of the filesystem because if you did it at the block level you'd end up checksumming all of the metadata, journal, etc which probably isn't wanted. Compression pretty much has to be done at the filesystem level because the block layer doesn't know how the filesystem will be doing delayed allocations, final on-disk layout, etc so we'll need to wait for something like reiser4 or ext4 to implement that.

Also the time it takes to fsck a file system is excesive and is a problem in enterprise environments for Linux. A multi-terrabyte file system can take days to properly recover.
Click to expand...

Yes and that's already being worked on separately.

XFS doesn't have this problem, generally, but it's so complicated that nobody realy knows how it works, at least not in Linux-land. SGI-land, maybe.
Click to expand...

I woudn't go that far, Christoph Hellwig is one that probably knows XFS as well as anyone and I don't think he ever worked for SGI. And I believe most of the SGI people who work on the Linux XFS port are ex-SGI people now.

Sun maintains that while the Unix model of layers is all fine and dandy they designed ZFS by refactoring the basic concepts of software raid and logical volume management with file systems. That the layered approach in this context is a aberration of evolution of storage rather then any sort of logical or correct division. That having it all in the file system context is actually a more correct approach and they actually belong together.
Click to expand...

Sun also maintained for years that GPL'ing Java would kill it but look what they're doing now.
 
Originally posted by: n0cmonkey
Announcing things like that is the best way to get Apple not to do them. Just ask ATI. 😛
Click to expand...

Guess you were right :Q
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top