From: Chris Browne <cbbrowne@acm.org>
Subject: Re: how to improve the performance of creating index
Date: Fri, 19 Oct 2007 08:52:05 -0400
Organization: cbbrowne Computing Inc
Lines: 34
Message-ID: <60hckngtca.fsf@dba2.int.libertyrms.com>
References: <01c601c81247$3b8c5650$b2a502f0$@ust.hk>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
User-Agent: Gnus/5.1008 (Gnus v5.10.8) XEmacs/21.4.20 (linux)
Cancel-Lock: sha1:d4q5LyFenA9+IR0QOlrKRDoF4H4=
To: pgsql-performance@postgresql.org

yinanli@cse.ust.hk ("Yinan Li") writes:
> 牋牋牋牋 I am trying to improve the performance of creating index.:p>
>
> 牋牋牋牋 I've set shared_buffers = 1024MB:p>
>
> 牋牋牋牋牋牋牋牋牋 Effective_cache_size = 1024MB:p>
>
> 牋牋牋牋牋牋牋牋牋 Work_mem = 1GB:p>
>
> 牋牋牋牋牋牋牋牋牋 Maintenance_work_mem=512MB:p>
>
> 牋牋牋牋 (I'm sure that the server process has received the SIGHUP signal):p>
>
>                      However, when create index, I found that the memory used by  Postgres is only 50MB.  And it is very
> slow. How to make it faster?:p>
>
>                      All helps are appreciated.:p>

Those values seem rather large, with the exception of the effective
cache size, which I would expect to be somewhat bigger, based on the
other values.

Note that the values for work_mem and maintenance_work_mem get used
each time something is sorted or maintained.  So if those values get
set high, this can pretty easily lead to scary amounts of swapping,
which would tend to lead to things getting "very slow."

You may want to do a census as to how much resources you have on the
server.  Knowing that would help people make more rational evaluations
of whether your parameters are sensible or not.
-- 
(reverse (concatenate 'string "moc.enworbbc" "@" "enworbbc"))
http://linuxdatabases.info/info/languages.html
All syllogisms have three parts, therefore this is not a syllogism.