Message-ID: <52B33069.1060203@optionshouse.com>
Date: Thu, 19 Dec 2013 11:44:09 -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: Dave Johansen <davejohansen@gmail.com>, pgsql-performance
 <pgsql-performance@postgresql.org>
Subject: Re: Unexpected pgbench result
References: 
 <CAAcYxUe11Qdx7h3RaRjuGGvp3oO7Wi-QNYrN=9TXBSv+P2u0hg@mail.gmail.com>
In-Reply-To: 
 <CAAcYxUe11Qdx7h3RaRjuGGvp3oO7Wi-QNYrN=9TXBSv+P2u0hg@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 11:00 AM, Dave Johansen wrote:

> When I run pgbench in "SELECT only" after doing "-i -s 2000" I get what
> appears to be good performance (60k-70k tps) but if I initialize a new
> database with "-i -s 4000" the tps drops to 4k-7k. Is this order of
> magnitude drop expected? Or is there something wrong with my hardware or
> database configuration that is causing this issue?

When you increase the size of the initialized pgbench tables, you 
increase the size on disk. My guess is that you doubled it so that the 
data no longer fits in memory. You can verify this yourself:

SELECT pg_size_pretty(sum(pg_database_size(oid))::bigint)
   from pg_database;

Any amount of memory you have that is smaller than that, will affect 
select performance. I can guarantee you will not get 60k-70k tps from 
anything short of an array of SSD devices or a PCIe NVRAM solution. Your 
'-s 2000' test was probably running mostly from memory, while the '-s 
4000' did not.

What you're seeing is the speed your records are being supplied from 
disk, plus whatever cache effects are there when records are read before 
they are flushed in favor of more recent data.

-- 
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