[OSM-dev] Postgres 8.4 performance trouble in osm2pgsql setup

[sly (sylvain letuffe)]

(sorry for breaking the thread)

Just a mail to give other results in experiments with postgres 8.4,

I'm in the progress of migrating from debian lenny (postgres 8.3) to squeeze 
(postgres 8.4)  and I'm of concerns about initial import speed with osm2pgsql 
and later diffs.

I was just going to do it blindly but after reading you'r mail I gave it a 
try, and haven't run into the speed regression you mentionned.

Test case :
wget http://download.geofabrik.de/osm/europe/france/alsace.osm.bz2
time osm2pgsql -C 1200 -s -S ./default.style -G -x -m -d gis ./alsace.osm.bz2
(...)

First server (squeeze postgres 8.4, osm2pgsql SVN version 0.69-exported)
real    4m33.505s

Second one (lenny postgres 8.3, osm2pgsql SVN version 0.69-exported)
real    5m4.108s

Unfortunately both server are not equivalent, so the test is not perfect, but 
I would say they are closed in configuration for such a test (First has a 
RAID0 array and 2GO RAM while the other as a RAID1 array and 32Go RAM)
Mermory doesn't seams relevant in this test case, so the 37s diff might just 
come from the RAID0.
But I haven't reproduced you'r "4 time slower"

The only tweak in the default configuration provided by both lenny and squeeze 
is :
shared_buffers = 128MB

-- 
sly
Sylvain Letuffe sylvain at letuffe.org
qui suis-je : http://slyserv.dyndns.org