Date: Thu, 12 May 2005 12:34:46 +0200 (MEST)
From: "Marc Mamin" <m.mamin@gmx.net>
To: pgsql-performance@postgresql.org
MIME-Version: 1.0
Subject: tuning Postgres for large data import (using Copy from)
Message-ID: <6428.1115894086@www73.gmx.net>
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit

Hello,


I'd like to tune Postgres for large data import (using Copy from).


here are a few steps already done:


1) use 3 different disks for:

	-1: source data
	-2: index tablespaces
	-3: data tablespaces
	
	
2) define all foreign keys as initially deferred


3) tune some parameters:


	max_connections =20
	shared_buffers =30000    
	work_mem = 8192      
	maintenance_work_mem = 32768    
	checkpoint_segments = 12

	(I also modified the kernel accordingly)


4) runs VACUUM regulary


The server runs RedHat and has 1GB RAM

In the production (which may run on a better server), I plan to: 

- import a few millions rows per day,
- keep up to ca 100 millions rows in the db
- delete older data


I've seen a few posting on hash/btree indexes, which say that hash index do
not work very well in Postgres;
currently, I only use btree indexes. Could I gain performances whole using
hash indexes as well ?

How does Postgres handle concurrent copy from on: same table / different
tables ?


I'd be glad on any further suggestion on how to further increase my
performances.


Marc


-- 
+++ Lassen Sie Ihren Gedanken freien Lauf... z.B. per FreeSMS +++
GMX bietet bis zu 100 FreeSMS/Monat: http://www.gmx.net/de/go/mail