Message-ID: <4E1DE28D.6020207@yahoo.com>
Date: Wed, 13 Jul 2011 11:23:09 -0700
From: lars <lhofhansl@yahoo.com>
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
	rv:1.9.2.18) Gecko/20110621 Fedora/3.1.11-1.fc14 Thunderbird/3.1.11
MIME-Version: 1.0
To: Kevin Grittner <Kevin.Grittner@wicourts.gov>
CC: Ivan Voras <ivoras@freebsd.org>, pgsql-performance@postgresql.org
Subject: Re: UPDATEDs slowing SELECTs in a fully cached		 database
References: <4E16479D.3050605@yahoo.com> <4E1A0CBC.6090009@yahoo.com>
	<4E1A31AB.9010201@postnewspapers.com.au>
	<1310359653.80687.YahooMailRC@web121705.mail.ne1.yahoo.com>
	<4E1AEDAB020000250003F1C3@gw.wicourts.gov>
	<4E1B41EA.90104@yahoo.com>
	<4E1B05CC020000250003F1E6@gw.wicourts.gov>
	<CAHyXU0wPhZsQ_pMzfUWW0O=t_+2uwF-CJQNw-nuaxkjbi2zPEQ@mail.gmail.com>
	<4E1B810D.4070404@yahoo.com> <4E1B90AF.1020408@yahoo.com>
	<ivho9v$6dp$1@dough.gmane.org> <4E1C9A05.7060005@yahoo.com>
	<4E1C5563020000250003F252@gw.wicourts.gov>
	<4E1CA8E9.6090108@yahoo.com> <4E1CB56E.7080801@yahoo.com>
	<4E1D697C020000250003F29A@gw.wicourts.gov>
In-Reply-To: <4E1D697C020000250003F29A@gw.wicourts.gov>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 07/13/2011 07:46 AM, Kevin Grittner wrote:
>
> I've mentioned this in a hand-wavy general sense, but I should have
> mentioned specifics ages ago: for a database where the active
> portion of the database is fully cached, it is best to set
> seq_page_cost and random_page_cost to the same value, somewhere in
> the 0.1 to 0.05 range.  (In your case I would use 0.05.)  In highly
> cached databases I have sometimes also found it necessary to
> increase cpu_tuple_cost.  (In your case I might try 0.02.)
>   
I've been doing that for other tests already (I didn't want to add too 
many variations here).
The Bitmap Heap scans through the table are only useful for spinning 
media and not the cache
(just to state the obvious).

As an aside: I found that queries in a cold database take almost twice 
as long when I make that change,
so for spinning media this is very important.

> Which raises an interesting question -- what happens to
> the timings if your SELECTs are done with synchronous_commit = off?

Just tried that...
In that case the WAL is still written (as seen via iostat), but not 
synchronously by the transaction (as seen by strace).

-- Lars