Message-ID: <4BFB837C.6010005@iathh.de>
Date: Tue, 25 May 2010 09:59:56 +0200
From: Joachim Worringen <joachim.worringen@iathh.de>
Organization: IAT GmbH
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 6.1; de;
	rv:1.9.1.9) Gecko/20100317 Lightning/1.0b1 Thunderbird/3.0.4
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: performance of temporary vs. regular tables
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit

Greetings,

in 
http://archives.postgresql.org/message-id/1056648218.7041.11.camel@jester, 
it is stated that the performance of temporary tables is "the same as a 
regular table but without
WAL on the table contents.".

I have a datamining-type application which makes heavy use of temporary 
tables to stage (potentially large amounts of) data between different 
operations. WAL is write-ahead

To effectively multi-thread this application, I (think I) need to switch 
from temporary to regular tables, because
- the concurrent threads need to use different connections, not cursors, 
to effectively operate concurrently
- temporary tables are not visible across connections (as they are 
across cursors of the same connection)

Thus, I wonder how much this will affect performance. Access on the 
temporary table is inserting (millions of) rows once in a single 
transaction, potentially update them all once within a single 
transaction, then select on them once or more.

Of course, eventually loosing the data in these tables is not a problem 
at all. The threads are synchronized above the SQL level.

Thanks for any input on how to maximize performance for this applicaiton.

  Joachim