An Apple official on Monday said Sun's open-source file system would not be in the next version of the Mac operating system, contradicting statements made last week by Sun's chief executive. During an interview with InformationWeek, Brian Croll, senior director of product marketing for the Mac OS, said, "ZFS is not happening," when asked whether Sun's Zettabyte File System would be in Leopard. Instead, Leopard would use Apple's current hierarchical file system, called HFS+. The Apple file system was first introduced in 1998 in Mac OS 8.0.
To view parent comment, click here.
To read all comments associated with this story, please click here.
Member since:2005-11-10
ZFS does not support per user or per group quota.
So for users you would have to make their own filesystem for each user to control space. A BIG headache. And plus not easy to adjust on the fly like normal quotas.
And for groups you have nothing?? No ability to quota by group at all.
So as a file server in an enterprise I would choose Windows or Linux (XFS or JFS) for that (Or Solaris with it's default file system)
And you are right about the ACL's. I have been reading in the open solaris fourms that the ACL implimentation could be better and has some issues.
(Looking for links to share)
Also does anyone know if ZFS is Journaling?
Also, XFS is pretty sweet, it can support Volumes up to 8 EIB (ZFS supports 16) That is what I tend to use on Linux if not Jfs.
Member since:2006-01-02
First you state the HFS+ is adequate and then you ask if ZFS is journaling. I think you could do with some extra reading (I don't really want to be insulting: here's a great source from the guy who did BFS for BeOS about the internal structures and merits of several commercial FSes: http://www.letterp.com/~dbg/practical-file-system-design.pdf. Page 37 covers HFS(+)).
In particular, HFS and HFS+ have a very radically different filesystem structure from every other filesystem. The only thing that comes close in structure is ReiserFS (not sure how the B-tree differs between 3 and 4, but I'm basing my words on what I've read about 4). HFS was perfect for the early macintosh, which was at the beginning of the desktop computer, but it has serious problems on a modern multi-program and multi-user system. Both the original and the plus versions have a complicated metadata structure that cannot be safely accessed and updated in parallel, and which does not make the most efficient use of the disk and the cache (after all, the early PCs would not waste much memory on something unnecessary like a disk cache). HFS+ added features, like journalling (which theoretically should alleviate some of the bottlenecks, depending on how it's implemented), but the complicated catalog structure is still there.
The referenced book does not include anything about ZFS-style systems because they did not really exist when BFS was being written. ZFS does not need journalling because the filesystem is basically just a journal itself. It's a new (and high-performance) implementation of the log-structured filesystem concept. Wikipedia has a good article on this type of FS. Doing this efficiently requires a lot of RAM for a buffer cache. This is true for NTFS as well, which is why we saw very different FSes like FAT32 and HFS in the early days.
ZFS is really a good way forward. But I don't think even OpenSolaris yet boots off of ZFS, so it'd be highly surprising if Apple beats them to that punch. And frankly, FS performance is not _that_ critical to desktop workstation users. ZFS might be good if Apple were to aim more solidly at the server market. Otherwise it's purely a marketing gimmick (that requires a lot of dev work), just to promote "the world's most advanced OS." To go for the server market, they'd probably want to just scrap the rest of the kernel, though, because XNU is not the greatest design, as far as I have read.
Member since:2006-10-23
Creating filesystems in ZFS is almost analogous to creating a directory. So creating a filesystem out of the ZFS pool really isn't a problem.
If you want to quota by group then you should look here http://www.opensolaris.org/os/community/zfs/faq/;jsessionid=97CB9.
AFAIK, ZFS isn't journaling in the normal sense, but it does ensure on disk consistency in the same way.
Member since:
2005-09-14
I think it will be really good for large client file systems. (People who have a LOT of music, MP3's and have external drives) And it's good for large file systems for databases etc. But ZFS doesn't support quotas and a lot of file permission features you find in other file systems.
So on file servers it sucks. :-(
Um...ZFS supports per-filesystem quotas and NFSv4 ACLs last I checked. Where's the problem?