Обсуждение:

Hi,
These last two days, I have some troubles with a very strange phenomena:
I have a 400 Mb database and a stored procedure written in perl which
call 14 millions times spi_exec_query (thanks to Tom to fix the memory
leak ;-) ).
On my laptop whith Centrino 1.6 GHz, 512 Mb RAM,
 - it is solved in 1h50' for Linux 2.6
 - it is solved in 1h37' for WXP Professionnal (<troll on> WXP better
tan Linux ;-) <troll off>)
On a Desktop with PIV 2.8 GHz,
 - it is solved in 3h30 for W2K
On a Desktop with PIV 1.8 GHz, two disks with data and index's on each disk
 - it is solved in 4h for W2K

I test CPU, memory performance on my laptop and it seems that the
performances are not perfect except for one single  test: String sort.

So, it seems that for my application (database in memory, 14 millions
of very small requests), Centrino (aka Pentium M) has a build-in
hardware to boost Postgres performance :-)
Any experience to confirm this fact ?

--
Jean-Max Reymond
CKR Solutions Open Source
Nice France
http://www.ckr-solutions.com

> So, it seems that for my application (database in memory, 14 millions
> of very small requests), Centrino (aka Pentium M) has a build-in
> hardware to boost Postgres performance :-)
> Any experience to confirm this fact ?

    On my Centrino, Python flies. This might be due to the very large
processor cache. Probably it is the same for perl. With two megabytes of
cache, sorting things that fit into the cache should be a lot faster too.
Maybe this explains it.

    Check this out :

http://www.anandtech.com/linux/showdoc.aspx?i=2308&p=5

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2129&p=11

    Bonus for Opteron lovers :

"The Dual Opteron 252's lead by 19% over the Quad Xeon 3.6 GHz 667MHz FSB"

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2397&p=12