Fsync Slashdot discussion

I am reading an interesting discussion about fsync() and disk flush on
Slashdot. The discussion starts about log-based file systems, then
moves to disk fsync about 20% into the discussion.

Look for:

Real issue is HARD DRIVE CACHEs

All this discussion relates to WAL and our use of fsync().

There is also a mention of PostgreSQL:

I just tried both about 3 weeks ago. I first tried reieserfs. Worked
fine until I tried to load a postgresql database. Never finished after
running all night--it realy thrashed the drive (I also run a RAID level
0 on the filesystem). Switched to xfs and everything ran great (db
loaded in a few hours). Been running xfs ever since. I'd like to see xfs
get put into the stock kernel along with reiserfs. It's possible my
problems with reiserfs have since been fixed.

Also, in reading the thread, it seems xfs is much more log-based than
Reiser, so we may only have WAL/fsync() performance problems on xfs.

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026