<br><br><div class="gmail_quote">On Thu, Sep 17, 2009 at 9:29 AM, Adam Tauno Williams <span dir="ltr"><<a href="mailto:awilliam@whitemice.org">awilliam@whitemice.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">> > Yes, these questions can be about as<br>
> > annoying to answer as "how are you?",<br>
> > or "What's up?" But helpful to some of<br>
> > us.<br>
> Just to throw in my 2 cents... ext3 tends to be the most widely<br>
> tested of the Linux filesystems, and has the most capable fsck. Other<br>
> filesystems may be a few seconds faster in one way or another, but if<br>
> you care about your data actually being there when you need it,<br>
> there's no safer choice on Linux than ext3.<br>
<br>
</div>We used to be an XFS shop, but with dir_index the single biggest<br>
performance plus of XFS was gone [at least for us], and ext3 is better<br>
supported with a richer tool chain. Of course xfs.fsck is much faster<br>
- it doesn't actually do anything [it admits that right in the man<br>
page]. The equivalent is xfs_check / xfs_repair which depending on the<br>
options and filesystem size can take some time to run.<br>
<br>
For productive systems I'm more comfortable with ext3 at this point -<br>
and it provides an upgrade path to ext4 which provides all the<br>
advantages of XFS [most notably: extents]<br>
<br>
If you have solid hardware and a UPS run ext3 as<br>
"noatime,data=writeback" and you'll probably get nearly equivalent<br>
performance. And an external journal helps, just as with XFS.<br>
<div><div></div><div class="h5"><br></div></div></blockquote><div><br>"Of course xfs.fsck is much faster<br>
- it doesn't actually do anything."<br><br>Quite a mouthful, and a reminder<br>of the potential pitfalls of comparisons.<br><br>Much harder is to ensure that one is<br>actually talking about equivalent value.<br>
<br>Interesting points.<br><br> -- Bob<br><br> </div></div>