[Top] [All Lists]

Re: extremely slow write performance plaintext

To: xfs@xxxxxxxxxxx
Subject: Re: extremely slow write performance plaintext
From: Mathieu AVILA <mathieu.avila@xxxxxxxxxxxxxxxx>
Date: Tue, 25 Jan 2011 10:48:35 +0100
In-reply-to: <201101250721.14716@xxxxxx>
References: <3205_1294953756_4D2F6D1C_3205_1943_1_4D2F6D1C.2060409@xxxxxxxxxxxxxxx> <4D30C7F3.3040105@xxxxxxxxxxxxxxxxx> <1779_1295360207_4D35A0CF_1779_747_1_4D35A0CF.9050503@xxxxxxxxxxxxxxx> <201101250721.14716@xxxxxx>
User-agent: Mozilla/5.0 (Windows; U; Windows NT 6.0; fr; rv: Gecko/20101207 Lightning/1.0b2 Thunderbird/3.1.7
Le 25/01/2011 07:21, Michael Monnerie a écrit :
On Dienstag, 18. Januar 2011 Cory Coager wrote:
Also, I want to stress that I am only having performance issues with
one logical volume, the others seem fine.
You're getting 40MB/s write speed and says that's fine? I get more 
performance from a single SATA desktop drive. Your setup seems to suck 
extremely somewhere.

It already starts with your dd speed. I started it on a virtualized VM 
on an old, overloaded server and get:

# dd if=/dev/zero of=test.dd bs=1k count=1M
1048576+0 Datensätze ein
1048576+0 Datensätze aus
1073741824 Bytes (1,1 GB) kopiert, 10,9992 s, 97,6 MB/s

I would say this is melted with cache hit, isn't it ? (although I agree it depends on the server's memory dedicated to FS cache)

A real cache hit on a real server would give me at least 500 MB/s (actually I get something around 1 GB/s)
When it starts to be disk-constrained, on a standard SATA disk, I get from 75 MB/s to 140 MB/s depending on the position of the AG I hit.

Mathieu Avila
IT & Integration Engineer

OpenCube Technologies http://www.opencubetech.com
Parc Technologique du Canal, 9 avenue de l'Europe
31520 Ramonville St Agne - FRANCE
Tel. : +33 (0) 561 285 606 - Fax : +33 (0) 561 285 635
<Prev in Thread] Current Thread [Next in Thread>