A question for the guys at SGI that are working so hard on this beast:
With the fundamental differences regarding implementation, in particular the
invasiveness of the XFS patches, and the talk now of possibly targeting 2.6
rather than 2.5, what's the general take on (1) capitulation to the Linus
and AC viewpoint about reducing the invasiveness by taking out some of the
code that dupes functionality in the kernel but does it differently and (2)
longevity of the XFS project in light of the industry's less than stellar
financial condition as a whole?
I'd certainly be bummed out serverely is the XFS project were to become a
victim of the tech industry depression, mostly because I just don't feel
right with ReiserFS (it's too new, and it's developers too egotistical for
my taste), reports on JFS are less than stellar (I certainly wouldn't trust
it for anything other than /tmp), reports on ext3 seem to indicate that it
may have some, err, performance issues (well, it's a retrofit), and what
does all that leave?
It's XFS or 6-12+ hour boot times (or just always restore from tape) in case
of a failure of power *and* UPS *and* auto-shutdown. Either way, without
XFS, a triple failure means you're down for most of a day.
I've got 120 Gig of disk on this box now, and it's only gonna get bigger.
XFS is the only choice in a situation like that, really.
And BSD doesn't really have any options for journaling that I'm aware of, so
switching to BSD isn't really an option. Nor is Solaris, 'cause I know I'm
not wealthy enough to buy Veritas, if it's even available for IA-32
architecture. For those of us with big disk space systems (and that group is
growing), Linux + XFS is the only real game in town right now AFAIK.
So, looking down the road a piece, after consulting your tarot cards,
calling your personal certified psychotic adviser, or reading the latest
chicken innards, what do you see?
--
Alan Eldridge
from std_disclaimer import *
|