On Sun, 13 Mar 2011, Holger Levsen wrote:
Hi,
On Samstag, 12. März 2011, Ben Hutchings wrote:
ext3 is inefficient for large files since it uses simple block bitmaps.
how inefficient is it? how large?
Depends on what you do with it, but some things like deleting a large file
(1-100GB) can take several seconds to minutes on ext3, yet be instant with
xfs. Same thing for mkfs for instance.
And xfs can be a bit smarter about file layout that might get a bit better
performance if you have several streaming reads/writes at the same time,
but that's minor optimisation.
At work we store a couple of PB on xfs, in filesystems of about 5-30TB
each. Filesizes here all the way up to 40GB.
What put me off from using xfs or reiserfs were quite frequent stories about
data loss with those filesystems (which I see until today). I rather have my
data served slower, but reliable.
Never really heard much about data loss on xfs that didn't have external
explanations - I even have very good experience with xfs_repair when
hardware failed badly. Reiserfs I wouldn't touch again thogh.
/Mattias Wadenstein
_______________________________________________
Debconf-video mailing list
[email protected]
http://lists.debconf.org/mailman/listinfo/debconf-video
