Message-ID: <51F9C19E.7010105@lab.ntt.co.jp>
Date: Thu, 01 Aug 2013 11:02:06 +0900
From: KONDO Mitsumasa <kondo.mitsumasa@lab.ntt.co.jp>
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64;
 rv:17.0) Gecko/20130620 Thunderbird/17.0.7
MIME-Version: 1.0
To: Tasos Petalas <tasos.petalas@upstreamsystems.com>
CC: pgsql-performance@postgresql.org
Subject: Re: PG performance issues related to storage I/O waits
References: <21cb334b6ca0c09c4faa60bdd46179eb@mail.gmail.com>
In-Reply-To: <21cb334b6ca0c09c4faa60bdd46179eb@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Hi Tasos,

(2013/07/31 20:58), Tasos Petalas wrote:
> We have set synchronous_commit to off but since almost all of application queries
> are using prepared statements we don't get any real benefit.

> Also please advise on any postgres.conf modification that could significantly
> affect storage load (WAL-checkpoint configuration etc.)
Please send us to your postgresql.conf and detail of your server (memory ,raid 
cache size and OS version).

I think your server's memory is big. You should set small dirty_background_ratio 
or dirty_background_byte. If your server's memory is big, PostgreSQL's checkpoint 
executes stupid disk-write.

By the way, I make new checkpoint scheduler and method now. If you could send 
more detail information which is like response time or load average, you will 
help me to make better my patch.

Best regards,
--
Mitsumasa KONDO
NTT Open Source Software Center


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance