Message-ID: <52B4574E.7040800@optionshouse.com>
Date: Fri, 20 Dec 2013 08:42:22 -0600
From: Shaun Thomas <sthomas@optionshouse.com>
User-Agent: Mozilla/5.0 (X11; Linux x86_64;
 rv:24.0) Gecko/20100101 Thunderbird/24.1.1
MIME-Version: 1.0
To: Sergey Konoplev <gray.ru@gmail.com>, Sev Zaslavsky <sevzas@gmail.com>
CC: "pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
Subject: Re: slow query - will CLUSTER help?
References: <52A9F2A2.4080901@gmail.com>
 <CAL_0b1ty0gS+XP9L=v+jsAT8KXMY5yAW2P0f2fjjjncdYLU3WA@mail.gmail.com>
 <52B35CF9.9080402@gmail.com>
 <CAL_0b1vGV0kWyVKJM=kOYonEXyt1O0a44STbDpknJKhbFyw_XQ@mail.gmail.com>
In-Reply-To: 
 <CAL_0b1vGV0kWyVKJM=kOYonEXyt1O0a44STbDpknJKhbFyw_XQ@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

On 12/19/2013 03:24 PM, Sergey Konoplev wrote:

> 2. You are limited with IO
> I would also suggest you to upgrade your storage in this case.

I think this is the case. If I recall correctly, his setup includes a 
single RAID-1 for everything, and he only has 32GB of RAM. In fact, the 
WAL traffic from those inserts alone are likely saturating the write IO, 
especially if it starts a checkpoint while the load is still going on. I 
wouldn't want to be around for that.

Even with a fairly selective index, just the fetches necessary to 
identify the rows and verify the data pages will choke a RAID-1 with 
almost every query. Any table with several hundred million rows is also 
too big to fit in cache if any significant portion of it is fetched on a 
regular basis. The cache turnover is probably extremely high, too.

That workload is just too high for a system of that description. It 
would be fine for a prototype, development, or possibly a QA system, but 
if that's intended to be a production resource, it needs more memory and IO.

Also since I can't see part of this conversation and it doesn't seem 
anyone else mentioned it, the WAL directory *must* be moved to a 
separate set of disks for a workload of this volume. The amount of 
writes here will constantly degrade read IO and further increase fetch 
times.

-- 
Shaun Thomas
OptionsHouse | 141 W. Jackson Blvd. | Suite 500 | Chicago IL, 60604
312-676-8870
sthomas@optionshouse.com

______________________________________________

See http://www.peak6.com/email_disclaimer/ for terms and conditions related to this email


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