Thread: Problem on Linux

Problem on Linux

From
"Molenda, Mark P"
Date:
I'm migrating tables from Solaris to Linux.  Other than Red-Hat moving the
directories a little bit I expected a close match on performance.

It seems that the SUN version (compiled from source) handled 13 million rows
with ( I know this is not efficient ) Select * from tableName;
The Linux box bumped me out of psql with the exact table structure but with
only 4 million rows doing the same select *.

I'm wondering if it is the startup of the system.  I had to issue a huge
nohup on the SUN and on linux it has a predefined /etc/rc.d/init.d script.

Linux postgres gurus'  -  I will have approx. 24 million rows per table
loaded each day.   I don't care if it takes 4 hours to complete an sql call
as long as it does complete.  I need to keep 90 days of data, so I'm
thinking of using a new (exact duplicate of) table each day for 90 days then

Truncate the oldest and start over.

Is there a better way?  And how do I tune the stock RedHat version?  I
really don't want to have to load it again from scratch (source).

P.S. I'm going from a 450MHz sparc to a 750MHz  compaq with similar amounts
of real and swap memory.

-Mark

Re: Problem on Linux

From
Nabil Sayegh
Date:
Am Don, 2003-05-22 um 16.33 schrieb Molenda, Mark P:

> Is there a better way?  And how do I tune the stock RedHat version?  I
> really don't want to have to load it again from scratch (source).

Did you try a: "VACUUM ANALYZE" ?
It calculates indexes etc. automatically (depending on your usage).

Also you could try "EXPLAIN <your query>" to get a hint how postgresql
handles it and maybe to improve the querys manually.

HTH
--
 e-Trolley Sayegh & John, Nabil Sayegh
 Tel.: 0700 etrolley /// 0700 38765539
 Fax.: +49 69 8299381-8
 PGP : www.e-trolley.de
--
 e-Trolley Sayegh & John, Nabil Sayegh
 Tel.: 0700 etrolley /// 0700 38765539
 Fax.: +49 69 8299381-8
 PGP : www.e-trolley.de


Re: Problem on Linux

From
Tom Lane
Date:
"Molenda, Mark P" <mark.molenda@eds.com> writes:
> It seems that the SUN version (compiled from source) handled 13 million rows
> with ( I know this is not efficient ) Select * from tableName;
> The Linux box bumped me out of psql with the exact table structure but with
> only 4 million rows doing the same select *.

Sounds like you have a low ulimit -d setting on the Linux system.

            regards, tom lane

Re: Problem on Linux

From
Nabil Sayegh
Date:
Am Don, 2003-05-22 um 16.33 schrieb Molenda, Mark P:

> Is there a better way?  And how do I tune the stock RedHat version?  I
> really don't want to have to load it again from scratch (source).

Did you try a: "VACUUM ANALYZE" ?
It calculates indexes etc. automatically (depending on your usage).

Also you could try "EXPLAIN <your query>" to get a hint how postgresql
handles it and maybe to improve the querys manually.

HTH
--
 e-Trolley Sayegh & John, Nabil Sayegh
 Tel.: 0700 etrolley /// 0700 38765539
 Fax.: +49 69 8299381-8
 PGP : www.e-trolley.de