Message-ID: <4C4B629C.70501@2ndquadrant.com>
Date: Sat, 24 Jul 2010 18:01:00 -0400
From: Greg Smith <greg@2ndquadrant.com>
User-Agent: Thunderbird 2.0.0.24 (X11/20100317)
MIME-Version: 1.0
To: Yeb Havinga <yebhavinga@gmail.com>
CC: pgsql-performance@postgresql.org
Subject: Re: Testing Sandforce SSD
References: <4C4A9452.9070100@gmail.com> <4C4B1F84.5030306@2ndquadrant.com>
	<4C4B43DA.90201@gmail.com> <4C4B4512.2060007@gmail.com>
	<4C4B5984.4010705@gmail.com>
In-Reply-To: <4C4B5984.4010705@gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Yeb Havinga wrote:
> Writes/s start low but quickly converge to a number in the range of 
> 1200 to 1800. The writes diskchecker does are 16kB writes. Making this 
> 4kB writes does not increase writes/s. 32kB seems a little less, 64kB 
> is about two third of initial writes/s and 128kB is half.

Let's turn that into MB/s numbers:

4k * 1200 = 4.7 MB/s
8k * 1200 = 9.4 MB/s
16k * 1200 = 18.75 MB/s
64kb * 1200 * 2/3 [800] = 37.5 MB/s
128kb * 1200 / 2 [600] = 75 MB/s

For comparison sake, a 7200 RPM drive running PostgreSQL will do <120 
commits/second without a BBWC, so at an 8K block size that's <1 MB/s.  
If you put a cache in the middle, I'm used to seeing about 5000 8K 
commits/second, which is around 40 MB/s.  So this is sitting right in 
the middle of those two.  Sequential writes with a commit after each one 
like this are basically the worst case for the SSD, so if it can provide 
reasonable performance on that I'd be happy.

-- 
Greg Smith  2ndQuadrant US  Baltimore, MD
PostgreSQL Training, Services and Support
greg@2ndQuadrant.com   www.2ndQuadrant.us