pg.ddx.io  pgsql-performance@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Josh Berkus <josh@agliodbs.com>
To: Jack Coates <jack@lyris.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-performance <pgsql-performance@postgresql.org>
Subject: Re: tuning questions
Date: Tue, 9 Dec 2003 09:35:04 -0800
Message-ID: <200312090935.04658.josh@agliodbs.com> (raw)
In-Reply-To: <1070989072.16079.19.camel@cletus.lyris.com>
References: <1070553983.6498.52.camel@cletus.lyris.com>
	<22185.1070911198@sss.pgh.pa.us>
	<1070989072.16079.19.camel@cletus.lyris.com>

Jack,

> Right, because re-architecture of a cross-platform query makes sense if
> performance is bad on all systems, but is questionable activity when
> performance is fine on some systems and lousy on others. Hence my
> statement that while SQL optimization is certainly something we want to
> do for across-the-board performance increase, I wanted to focus on other
> issues for troubleshooting this problem. I will be back to ask about
> data access models later :-)

Yes, but an EXPLAIN ANALYZE will also help show issues like sorts running out 
of memory, etc.   Really, we don't currently have enough information to do 
more than speculate; it's like trying to repair a car engine wearing a 
blindfold.

Particularly since it's possible that there are only 1 or 2 "bad queries" 
which are messing everything else up.

For that matter, it would really help to know:
-- How many simulatneous connections are running update queries during this 
process?
-- How about some sample VACUUM VERBOSE results for the intra-process vacuums?

> I ended up going back to a default postgresql.conf and reapplying the
> various tunings one-by-one. Turns out that while setting fsync = false
> had little effect on the slow IDE box, it had a drastic effect on this
> faster SCSI box and performance is quite acceptable now (aside from the
> expected falloff of about 30% after the first twenty minutes, which I
> believe comes from growing and shrinking tables without vacuumdb
> --analyzing).

Well, that brings 2 things immediately to mind:
1) That may improve performance, but it does mean that if your machine loses 
power you *will* be restoring from backup.   It's risky to do.

2) Your IDE system has write-caching enabled.   Once again, this is a nice 
performmance boost, if you don't mind database corruption in a power-out.


-- 
Josh Berkus
Aglio Database Solutions
San Francisco



view thread (32+ messages)

Message-ID: <200312090935.04658.josh@agliodbs.com>
Permalink:  ../200312090935.04658.josh@agliodbs.com/
Also on:    postgresql.org/message-id/200312090935.04658.josh@agliodbs.com

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-performance@postgresql.org
  Cc: josh@agliodbs.com, jack@lyris.com, tgl@sss.pgh.pa.us
  Subject: Re: tuning questions
  In-Reply-To: <200312090935.04658.josh@agliodbs.com>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox