David Lang wrote:
On Tue, 14 Mar 2006, Chris wrote:

The only other thing I can see is the old server is ext2:
/dev/hda4 on / type ext2 (rw,errors=remount-ro)

the new one is ext3:
/dev/hda2 on / type ext3 (rw)


this is actually a fairly significant difference.

with ext3 most of your data actually gets written twice, once to the journal and a second time to the spot on the disk it's actually going to live.

in addition there are significant differences in how things are arranged on disk between the two filesystems, (overridable at mount, but only changes future new files). the ext3 layout is supposed to be better for a general purpose filesystem, but I've found common cases (lots of files and directories) where it's significantly slower, and I think postgres will fall into those layouts.

try makeing a xfs filesystem for your postgres data and see what sort of performance you get on it.

Interesting.

To be honest I think I'm just lucky with my really old server. I can't see any particular tweaks in regards to drives or anything else. I have another server running postgres 7.4.something and it's as slow as the 8.1 system.

#1 is running 2.4.x kernel - pg 7.4 (debian package) - good performance. ext2.

#2 is running 2.2.x kernel (I know I know).. - pg 7.4 (debian package) - reasonable performance. ext2.

#3 is running 2.6.x kernel - pg 8.1 (fedora package) - reasonable performance. ext3.

--
Postgresql & php tutorials
http://www.designmagick.com/

---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
      choose an index scan if your joining column's datatypes do not
      match

Reply via email to