Welcome to the MacNN Forums.

C.A.T.S. CEO · Jun 26, 2007, 02:28 AM

Linky

Awesome.

krove · Jun 26, 2007, 03:39 AM

What we really want to know, then, is whether it will be bootable.

C.A.T.S. CEO · Jun 26, 2007, 04:22 AM

Originally Posted by krove

What we really want to know, then, is whether it will be bootable.

I doubt it, it seems like it will be read/write/format in 10.5, maybe bootable in 10.6.

Peter · Jun 26, 2007, 04:53 AM

did you just pluck 10.6 out of the air?

OreoCookie · Jun 26, 2007, 12:04 PM

You forgot to add `Good news, everyone'

C.A.T.S. CEO · Jun 26, 2007, 04:25 PM

Originally Posted by Peter

did you just pluck 10.6 out of the air?

Yes, its all rumor right now, really.

jasong · Jun 27, 2007, 06:11 PM

Actually I heard it won't be until 10.7.3

C.A.T.S. CEO · 02:46 AM

My eyes are failing me....

MindFad · Jun 27, 2007, 11:01 PM

GnOm · Jun 28, 2007, 02:21 AM

Go out and buy some RAM

You sure all know the wikipedia article with the known issues, in the Opensolaris mailing list there are some quite interesting discussions too.
I'd rather have them give it some time to sort the quirks out.

cheers.

Mrjinglesusa · Jun 28, 2007, 08:54 AM

Can someone (briefly) explain why ZFS is a big deal? Seems a lot of people were disappointed it wasn't announced.

TETENAL · Jun 28, 2007, 10:00 AM

Originally Posted by Mrjinglesusa

Can someone (briefly) explain why ZFS is a big deal? Seems a lot of people were disappointed it wasn't announced.

It's a big deal because it's a file system that is brand new in design, and when you read the Wikipedia-article nobody really understands anything, except that there are a lot of really big numbers with exponents and stuff. So it must be great.

C.A.T.S. CEO · 09:53 PM

Originally Posted by Mrjinglesusa

Can someone (briefly) explain why ZFS is a big deal? Seems a lot of people were disappointed it wasn't announced.

It, supposedly, handles large disks better, is better for RAIDing, and some people said that you could fast OS switch like fast user switching in OS X.

Chuckit · Jun 28, 2007, 10:18 AM

Originally Posted by Mrjinglesusa

Can someone (briefly) explain why ZFS is a big deal? Seems a lot of people were disappointed it wasn't announced.

It's just a very good file system. It's fast, it has very good data integrity and it even supports Time Machine-like functionality at the file system level. If you care a lot about the file system (frankly, I don't), then you'll want ZFS.

besson3c · Jun 28, 2007, 10:26 AM

Originally Posted by C.A.T.S. CEO

It handles large disks better, is better for RAIDing, and some people said that you could fast OS switch like fast user switching in OS X.

Where does it say that in the Wikipedia article?

besson3c · Jun 28, 2007, 10:29 AM

Originally Posted by Chuckit

It's just a very good file system. It's fast, it has very good data integrity and it even supports Time Machine-like functionality at the file system level. If you care a lot about the file system (frankly, I don't), then you'll want ZFS.

I agree, the average home user won't even know that it is running.

C.A.T.S. CEO · Jun 28, 2007, 10:37 AM

Originally Posted by besson3c

Where does it say that in the Wikipedia article?

I donno, thats what I have heard, not from the Wikipedia article. I have no idea if its TRUE if thats what you're getting at....

MacosNerd · Jun 28, 2007, 10:49 AM

Originally Posted by C.A.T.S. CEO

Linky

Awesome.

Calling a file system awesome seems weird, even on a geek board

besson3c · Jun 28, 2007, 11:03 AM

Originally Posted by C.A.T.S. CEO

I donno, thats what I have heard, not from the Wikipedia article. I have no idea if its TRUE if thats what you're getting at....

That's why it's good to add an "I believe" or "I think" qualifier to claims you aren't certain of, especially as long as you insist on being rude to people whom you perceive as knowing less than you, like you were doing in that other thread.

Peter · Jun 28, 2007, 12:50 PM

Just because a file system is new doesn't make it good.
HFS+ is about 10 years old and its bloody good.

Catfish_Man · 01:37 PM

Right, ZFS is good because it's extremely well done according to everything I've read (and I've spent a nontrivial amount of time specifically googling for negative reviews), not because it's new.

<edit>
If you want some substance as to what's good about it, here's a brief list of some of the bits I care about:
1) Data checksumming. Improves reliability.
2) RAID-Z; most of the benefit of RAID-5 without the expensive hardware controller
3) IO prioritization and batching. Improves performance, particularly when under heavy IO load (which is an area OSX is terrible at right now)
4) Snapshots. Should allow time machine to be implemented in a less disk-intensive way.
5) Compression. Saves space, and can theoretically speed things up by trading CPU time for IO reduction.
</edit>

besson3c · Jun 28, 2007, 01:39 PM

Originally Posted by Peter

Just because a file system is new doesn't make it good.
HFS+ is about 10 years old and its bloody good.

What makes you say that it is bloody good?

besson3c · Jun 28, 2007, 01:41 PM

Originally Posted by Catfish_Man

5) Compression. Saves space, and can theoretically speed things up by trading CPU time for IO reduction.

And this feature is very compelling for running certain kinds of servers that are I/O bound more than they are CPU bound.

Peter · Jun 28, 2007, 04:02 PM

Originally Posted by besson3c

What makes you say that it is bloody good?

uhmm, it works and is tried and tested?

Catfish_Man · Jun 28, 2007, 04:09 PM

Originally Posted by Peter

uhmm, it works and is tried and tested?

People undervalue this to a surprising degree, I've found.

besson3c · Jun 28, 2007, 04:18 PM

Originally Posted by Peter

uhmm, it works and is tried and tested?

Rixstep

Thus the search for a file must always start with searching for the CNID of the parent folder - and if it has a parent, finding that folder's CNID, and so forth, until one arrives at the Unix 'root' directory ('/').

Thus all files on disk end up being sorted in the above order in the B*-tree. The effect is a sort by sub-folder, so listing a folder's contents is a matter of walking the forward and backward links between index nodes at the right level.

The implication of this 'caveat' is that a disk access in Mac OS X requires at least two searches through the catalog file tree: the first to determine the parent folder's CNID (this can be recursive), and the second (or final) to find the file once the CNID of the parent folder can be concatenated with the file name.

Word has it the NextStep people in Cupertino hate HFS+. How very strange.

NextStep of course used UFS - not HFS+.

Had Apple used UFS instead of HFS+, OS X might have been released several years sooner, Mac hardware would today last a lot longer, and Mac file systems would crash a lot less often.

Chuckit · Jun 28, 2007, 04:21 PM

Originally Posted by besson3c

And this feature is very compelling for running certain kinds of servers that are I/O bound more than they are CPU bound.

I would guess most computer use outside of science applications is mainly IO-bound rather than CPU-bound. For instance, on OS X, launch an app that requires several bounces and monitor your CPU usage — for most apps, that time is being spent mostly on IO rather than actual processing.

besson3c · Jun 28, 2007, 04:21 PM

Originally Posted by Chuckit

I would guess most computer use outside of science applications is mainly IO-bound rather than CPU-bound. For instance, on OS X, launch an app that requires several bounces and monitor your CPU usage — for most apps, that time is all being spent on IO rather than actual processing.

and some computer use is also memory bound.

CharlesS · Jun 28, 2007, 04:30 PM

Disclaimer: I don't know much about file systems, and can't affirm or refute anything that this guy says.

With that said, here's a link that attempts to explain the shortcomings of HFS+: Rixstep

edit: whoops, i seem to have been beaten to the punch with the exact same link. Odd.

Catfish_Man · Jun 28, 2007, 04:39 PM

I'm not familiar enough with filesystems to say much about the contents of that link, but I am definitely inclined to distrust anything from rixstep.

Also, it's pretty clear that whatever Apple inherited from NeXT in terms of UFS support is no longer cutting it; OSX's UFS implementation has consistently shown itself to be slower and more problematic than its HFSX implementation. I've heard some of the other UFS implementations are much better, but I don't know how true that is.

OreoCookie · Jun 28, 2007, 04:58 PM

FreeBSD uses UFS (now: UFS2) and the implementation supposed to be very good. They use Soft Updates instead of journaling to maintain data integrity (as far as I understand, this is similar to Sun's approach in ZFS).

On OS X UFS has never been the filesystem of choice, it was a step child that even broke some apps in the beginning and I doubt Apple has invested a lot of manpower into it. Even if Apple doesn't use ZFS by default now, it's clear that it will eventually become the filesystem of choice: it's mature (thanks to Sun's engineers), is lightyears ahead of HFS+ in terms of features and future-proof. I don't think Apple's more conservative approach should fool you: I think OS X has a very rich and solid foundation with all its Core services and APIs, ZFS is just another piece of the mosaic.

besson3c · Jun 28, 2007, 05:01 PM

Originally Posted by CharlesS

edit: whoops, i seem to have been beaten to the punch with the exact same link. Odd.

Nyah nyah!

Great minds think alike

besson3c · Jun 28, 2007, 05:03 PM

Originally Posted by Catfish_Man

I'm not familiar enough with filesystems to say much about the contents of that link, but I am definitely inclined to distrust anything from rixstep.

Wikipedia isn't, this article is linked to from Wikipedia.

Also, it's pretty clear that whatever Apple inherited from NeXT in terms of UFS support is no longer cutting it; OSX's UFS implementation has consistently shown itself to be slower and more problematic than its HFSX implementation. I've heard some of the other UFS implementations are much better, but I don't know how true that is.

There are several reasons why Apple went with HFS+ over UFS, probably many of them relating to Carbon and Classic. However, there is nothing wrong with UFS. It remains the default file systems on the BSD operating systems, including FreeBSD.

Catfish_Man · Jun 28, 2007, 05:19 PM

Originally Posted by besson3c

Wikipedia isn't, this article is linked to from Wikipedia.

Wikipedia isn't exactly an authoritative source either

That said, I'm not really contesting the information here.

There are several reasons why Apple went with HFS+ over UFS, probably many of them relating to Carbon and Classic. However, there is nothing wrong with UFS. It remains the default file systems on the BSD operating systems, including FreeBSD.

Which is why I was very careful to say "OSX's UFS implementation", rather than "UFS".

besson3c · Jun 28, 2007, 05:28 PM

Originally Posted by Catfish_Man

Wikipedia isn't exactly an authoritative source either

Oh?

besson3c · Jun 28, 2007, 05:33 PM

Originally Posted by OreoCookie

FreeBSD uses UFS (now: UFS2) and the implementation supposed to be very good. They use Soft Updates instead of journaling to maintain data integrity (as far as I understand, this is similar to Sun's approach in ZFS).

I believe that soft updates was added to FBSD 5.x, and the metadata journaling in HFS+ in 10.3?

If so, my point is that I don't think this was in the mix when Apple originally chose HFS+ over UFS.

OreoCookie · Jun 28, 2007, 05:54 PM

Yes, it was added in 5.0, as was fsck in multi-user mode (very useful addition). I'm not saying that UFS was somehow superior when OS X first came out … I rather believe that Apple chose HFS+ for backwards compatibility: using UFS as filesystem broke certain Carbon apps, so the choice was easy. And Apple has never seen a convincing reason to switch to UFS.

I was just trying to answer that there are fast implementations of UFS out there with modern features. I suspect FreeBSD's implementation of UFS with Soft Updates is faster than HFS+, but that this doesn't really have an impact for the average user. I don't think choosing HFS+ over UFS was a bad thing™.

besson3c · Jun 28, 2007, 07:00 PM

Originally Posted by OreoCookie

Yes, it was added in 5.0, as was fsck in multi-user mode (very useful addition). I'm not saying that UFS was somehow superior when OS X first came out … I rather believe that Apple chose HFS+ for backwards compatibility: using UFS as filesystem broke certain Carbon apps, so the choice was easy. And Apple has never seen a convincing reason to switch to UFS.

I was just trying to answer that there are fast implementations of UFS out there with modern features. I suspect FreeBSD's implementation of UFS with Soft Updates is faster than HFS+, but that this doesn't really have an impact for the average user. I don't think choosing HFS+ over UFS was a bad thing™.

Nor do I, and I'm sure that UFS would have broken Classic apps too.

My main point was simply that HFS+ has simply been adequate for Apple, but not ideal.

OreoCookie · Jun 28, 2007, 07:07 PM

Then we are in absolute agreement. (BTW, it's a fact that UFS broke Carbon apps in the early times of OS X which, I think, was connected to case sensitivity vs. case preservation, it's been a long time.)

TETENAL · Jun 28, 2007, 07:47 PM

Originally Posted by besson3c

Had Apple used UFS instead of HFS+, OS X might have been released several years sooner, Mac hardware would today last a lot longer, and Mac file systems would crash a lot less often.

Mac OS X was first released in 2000 as public beta. "Several years sooner" would be 1997, right after Apple acquired NeXT. Had they released Mac OS X then, they would have released NextStep as is. I'm sure some of the NeXT-guys would have loved this, but that would have hardly qualified as "Mac" OS.
And there was more work to do than just switching to HFS+. And more serious work. Switching from Display PS to the PDF-based Quartz is just one example. I doubt that HFS+ was a reason for a several years delay.

And since when does the filesystem make the hardware "not last as long". That's the first time I hear something like that. Sounds like bullshit to me. I also never had the filesystem "crash" as far as I could tell. I don't even know how such a thing would look like.

Brass · 08:51 PM

I concur... I've never come across any evidence that HFS+ causes any instability at all. Sounds like this rixstep has a chip on his shoulder and is grasping at straws.

Art Vandelay · Jun 28, 2007, 08:25 PM

Mac OS X Server 1.0, which was essentially Rhapsody, had HFS+ support in 1999. Mac OS X 10.0 was delayed for other reasons as noted above, not for HFS+ support.

Chuckit · Jun 28, 2007, 08:25 PM

Originally Posted by Brass

I concur... I've never come across any evidence that HFS+ causes any instability at all. Sounds like this rixstep has a chip on his shoulder and it grasping at straws.

Rob Rix is a huge Apple troll. He hates Apple and the Macintosh. He invents totally bizarre drama just so he'll have something to eviscerate Apple for. It seems like the only reason he'll even deign to look at OS X is because it's based on NeXTstep and FreeBSD, which he views as good Unix systems.

(NOTE: I don't regularly read Rob's site, but I've run into him before. I came up with those just by searching a couple of keywords on Google and randomly clicking.)

besson3c · Jun 28, 2007, 08:48 PM

What Rob Rix says is not untrue. If two searches through the catalog tree are required in HFS+, this means that using HFS+ is more I/O intensive. If it is more I/O intensive, hard disks will fail sooner.

Of course, there are a ton of other variables. In order to prove this, you'd have to do exactly the same stuff on identical hardware until the drive failed, and even then there are no guarantees that the experiment would work out as predicted since it is possible that one drive will fail before the other just because.

Secondly, as far as stability goes, if there is no support in the file system to maintain integrity of data, corrupt data can lead to crashes. As of 10.3 HFS+ added support for journaling, but this is metadata-only journaling. From the Wikipedia article about metadata-only journaling:

Journaling file system - Wikipedia, the free encyclopedia

Journaling can have a severe impact on performance because it requires that all data be written twice. Metadata-only journaling is a compromise between reliability and performance that stores only changes to file metadata in the journal. This still ensures that the file system can recover quickly when next mounted, but leaves an opportunity for data corruption because unjournaled file data and journaled metadata can fall out of sync.

Sorry guys. Just because you want this guy to be wrong doesn't make it so.

Brass · Jun 28, 2007, 08:55 PM

Originally Posted by besson3c

What Rob Rix says is not untrue. If two searches through the catalog tree are required in HFS+, this means that using HFS+ is more I/O intensive. If it is more I/O intensive, hard disks will fail sooner.

Of course, there are a ton of other variables. In order to prove this, you'd have to do exactly the same stuff on identical hardware until the drive failed, and even then there are no guarantees that the experiment would work out as predicted since it is possible that one drive will fail before the other just because.

Secondly, as far as stability goes, if there is no support in the file system to maintain integrity of data, corrupt data can lead to crashes. As of 10.3 HFS+ added support for journaling, but this is metadata-only journaling. From the Wikipedia article about metadata-only journaling:

Journaling file system - Wikipedia, the free encyclopedia

Sorry guys. Just because you want this guy to be wrong doesn't make it so.

... and yet, in the real world, working with many Macs for a long time, I've never experienced such problems, nor known anyone who has. Sounds like the problems are all either theoretical, or so minor or trivial so as not to be worth worrying about.

Most of my work is on Solaris systems using UFS, mind you, so I'm not a great fan of HFS+. It just seems to me that the gripes against it's use in Macs OS X are largely unfounded.

Horsepoo!!! · Jun 28, 2007, 09:16 PM

Originally Posted by Peter

uhmm, it works and is tried and tested?

By your definition FAT is good...and so is MFS. Both work...and both were tried and tested. Also, MS-DOS is good.

CharlesS · Jun 28, 2007, 09:58 PM

Originally Posted by Brass

... and yet, in the real world, working with many Macs for a long time, I've never experienced such problems, nor known anyone who has.

I have.

I've also seen plenty of reports of that happening on boards just like this one.

HFS+ gets corrupted. A lot. Why do you think DiskWarrior is such a popular app? I've had to use it to save my ass way more times than I should have had to. And that's under fairly normal usage. God forbid if you ever let the disk get full. Every person I've ever seen who had an almost-full HFS+ disk had a seriously messed-up directory, sometimes including thousands of overlapped files, something which AFAIK has been a known issue of HFS+ since the OS 9 days.

Those are all software problems, of course, not hardware, but as besson3c said, if it's creating more disk activity, it's going to wear out the mechanism faster. Have I ever had this happen to me? Well, I've had hard disks fail a few times. Of course, it's hard to determine exactly the cause of something like that. This is the problem with relying on anecdotal evidence, so I'll just go with heavier disk usage == shorter disk life, which makes sense.

Brass · 10:14 PM

Originally Posted by CharlesS

I have.

I've also seen plenty of reports of that happening on boards just like this one.

HFS+ gets corrupted. A lot. Why do you think DiskWarrior is such a popular app? I've had to use it to save my ass way more times than I should have had to. And that's under fairly normal usage. God forbid if you ever let the disk get full. Every person I've ever seen who had an almost-full HFS+ disk had a seriously messed-up directory, sometimes including thousands of overlapped files, something which AFAIK has been a known issue of HFS+ since the OS 9 days.

Well, I guess I'm just lucky then.

I've never had a need for those sort of tools, so I suppose I should be thankful. I had a client who thought he needed such a tool once, and it totally destroyed all his data (yes, it was a Norton tool).

Those are all software problems, of course, not hardware, but as besson3c said, if it's creating more disk activity, it's going to wear out the mechanism faster. Have I ever had this happen to me? Well, I've had hard disks fail a few times. Of course, it's hard to determine exactly the cause of something like that. This is the problem with relying on anecdotal evidence, so I'll just go with heavier disk usage == shorter disk life, which makes sense.

I've certainly not seen many HFS+ formatted disks failing when compared to any other format. They all get a lot of activity, they all fail now and then. I'd like to see some evidence (eg statistics) that HFS+ disks fail more often than UFS, before I go believing it. Some HFS+ disks having failed for some people isn't very good evidence of anything at all, as this happens for any file system.

Incidentally, does anyone know which components of an HDD are most likely to wear out? I would have though it would be the spindle or bearings. Since HDDs are usually spinning all the time (no matter the file system or usage pattern) unless specifically instructed to spin down (only when no usage at all - for any filesystem), then I would have expected this type of wear to be the same for disks with all file systems.

Of course this is all just guessing... I don't really have any experience or particular knowledge in this area. It's not as though when we return our many broken hard disks (all UFS by the way

) they tell us what the fault was later.

besson3c · Jun 28, 2007, 10:10 PM

Why is it a big surprise that HFS+ is not ideal? The file system has evolved, we've learned more about what kinds of things can be done with it and what works optimally. There are many other file systems that have come and gone, there have been many evolutions and iterations. We've patched on larger block support, journaling, xattr, and possibly other stuff, but at the heart of HFS+ is still HFS. You can only add on so much before it is better to make a clean break and start anew.

There are logistical considerations in moving to another file system: backwards compatibility, getting developer buy in if any changes are necessary, etc. Because there hasn't been a way to address these sorts of concerns in a satisfactory way doesn't mean that HFS is a wonderful file system.

besson3c · Jun 28, 2007, 10:14 PM

Originally Posted by Brass

I've certainly not seen many HFS+ formatted disks failing when compared to any other format. They all get a lot of activity, they all fail now and then. I'd like to see some evidence (eg statistics) that HFS+ disks fail more often than UFS, before I go believing it. Some HFS+ disks having failed for some people isn't very good evidence of anything at all, as this happens for any file system.

Like I said before, proving or disproving this would be next to impossible. All we have is theory to go on, but this theory seems completely sound and reasoned to me, and many others.

In theory a Unix operating system is more stable than a Windows operating system, but you cannot always count on consistent results that prove this unless you initiate a carefully performed experiment - one that we can't really do here.