Received: from argh.demon.co.uk (IDENT:grim@argh.demon.co.uk [193.237.6.55])
	by hub.org (8.9.3/8.9.3) with ESMTP id VAA88245
	for <pgsql-general@postgreSQL.org>;
	Sun, 11 Jul 1999 21:33:14 -0400 (EDT)
	(envelope-from grim@argh.demon.co.uk)
Received: (from grim@localhost) by argh.demon.co.uk (8.8.7/8.8.7) id CAA07664
	for pgsql-general@postgreSQL.org; Mon, 12 Jul 1999 02:30:58 +0100
From: M Simms <grim@argh.demon.co.uk>
Message-Id: <199907120130.CAA07664@argh.demon.co.uk>
Subject: A few questions
To: pgsql-general@postgreSQL.org
Date: Mon, 12 Jul 1999 02:30:58 +0100 (BST)
Content-Type: text

Hi

I asked these questions a couple of weeks ago and got no response whatsoever
so I am going to try again.

I have just installed 6.5, and there are some things I cannot find in
the documentation.

1 ) When I use temp tables, is there a way to instruct postgresql to
   keep these in memory rather than on disc, for faster access, or
   does it do this anyway with temp tables

2 ) Is there an optimal amount of updates and inserts to perform
   before vacuuming a database, some kind of formula based on inserts
   and updates that indicates when a vacuum would be most
   beneficial. I realise there cannot be an absolute rule for this,
   but a guideline would help, as I dont know if I will need to
   vacuum more than once a day on a busy database.

3 ) Is there a way to instruct postgresql to perform a query at a
   lower priority, such as daily maintainence operations, so that
   these jobs do not impact on the interactive actions. I realise
   I can renice a process that is making calls to the database, but
   that doesnt have any effect on the backend spawned by the
   postmaster when I connect to it.
   If there is no such functionality, would people be interested in it
   if I was to code it and release it back to the main source tree?

4 ) Is there an optimal ratio between the number of backends and the
   number of shared memory buffers. I realise there is a minimum of
   1:2 but do more shared memory buffers increase performance in some
   areas, or would the extra overhead of managing the buffers make the
   increase pointless.

5 ) The final question (I promise) is that if I have a large number of
   inserts that I generate dynamically, is it quicker for me to
   perform these inserts one by one (maybee 10,000 of them at a time)
   or would it be faster and less CPU intensive to generate a text
   file instead and then read this in via a single copy command.
   This file at times may be over 100,000 entries, so would I be
   better to split it to a maximum number of transactions if I
   take the route of the copy command?

Thanks in advance, and I hope that this time someone will be able
to answer some or all of these questions.

					M Simms

PS. Appologies to the person that receives this twice, I hit reply
instead of group reply to your mail to this list, and so you got
yourown personal copy {:-)