XFS is a high-performance journaling file system. It provides quick recovery after a crash, fast transactions, high scalability, and excellent bandwidth. This new release adds support for multiple filesystem block sizes and striped logs, plus plenty of bugfixes. SGI also provides an OS installer for Red Hat users who would like to use XFS instead.
XFS 1.2 Released
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
While XFS is still my filesystem of choice amoung my Linux systems, I have to question the “high performance” tagline. When I say this, I am, of course, talking about the Linux port and not its performance under Irix.
From my own experiences and benchmarks [ http://www.uwsg.iu.edu/hypermail/linux/kernel/0301.3/1303.html ], I’ve seen that XFS tends to be somewhat slower than some of the other filesystem offerings, namely ext3 and reiserfs, both of which are native to the Linux kernel.
I’ve also run into some frustration with the XFS tools. For example, xfs_repair can’t be done on a filesystem which is mounted read only, which tends to preclude running xfs_repair on the root filesystem.
There have also been past issues with bugs in XFS, such as when a metadata operation would be journaled before the write cache is flushed, so if you modified a file and something caused your system to reboot before a sync, the file you modified would be filled with zeroes when the system came back up. This has since been fixed, but doesn’t bode well for XFS’s reputation.
Overall, I’d say the page buffer abstraction added by SGI to overcome incompatibilities in the Linux XFS design has added a great number of problems which affect the performance and reliability of XFS. I can’t really say why I continue to use XFS… perhaps it was just blind Irix fandom.
Overall, I’d say the page buffer abstraction added by SGI to overcome incompatibilities in the Linux XFS design has added a great number of problems which affect the performance and reliability of XFS.
Please note: I meant “overcome compatibilities in the Linux VFS design”, not XFS design.
Interesting. Personally, I’ve found XFS’s performance to be in line with other benchmarks that have been conducted (which show great large file performance, mediocre small file performance). I wonder, what was the average file sizes in those benchmarks you did?
PS> XFS’s tools are unmatched by any other Linux filesystem. I’ve personally used xfs_repair a couple of times to fix, among other things, an MBR that I accidentally hosed.
all im saying is “Nice”(tm)
i bet redhat isnt too pleased though
I tried out XFS a few months ago. I found that XFS would pause for an unnecessarily long period of time when running ls on large directories, especially if I hadn’t accessed that directory for a few minutes or hours. I never experienced that behavior with ext2, and I don’t have that problem with reiserfs now.
I’ve never experienced any slowdown with ls with XFS, but it seems to slow me down when I log in. With ext2 I would put in my user name and password and be right at bash. With XFS it runs fortune and then pauses for a second for seeminly no reason, then I hit bash. It may also have to do with other system changes(my home directory is now now on ide raid), so I’m not sure. Is there any way XFS could be causing the slow down?
Does this resolve any of the patch conflict problems Gentoo was reporting?
I wonder, what was the average file sizes in those benchmarks you did?
I think my wording was a bit misleading. Perhaps if I had worded it “From benchmarks and my own experiences” it would’ve been more understandable. Anyway, those benchmarks are not my own.
PS> XFS’s tools are unmatched by any other Linux filesystem. I’ve personally used xfs_repair a couple of times to fix, among other things, an MBR that I accidentally hosed.
I’ll say this, even with the foibles I mentioned the XFS tools are far superior to the ReiserFS ones. I remember after experiencing some filesystem corruption (due to bugs in the ReiserFS implementation as far as I can gather, the underlying filesystem was fine) the ReiserFS tools recommended that I rebuild the b+tree, with the caveat that rebuilding the b+tree could destroy my entire filesystem and all my data. “Lovely,” I thought.
> the XFS tools are far superior to the ReiserFS ones
I beleive this, I lost a partition and the ReiserFS tools were .. not very usefull.
Since then I switched to ext3FS: do you know wether XFS or ext3FS has better tools for recovering from a corruption?
…are toughted to be top notch.
It sounds like your slowdown might be XFS related. It’s pretty similar to my own experience. I use a graphical autologin for my machine, anyway, so I wouldn’t notice at login.