This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


Re: [Xen-users] How to get rid of IO-Bottleneck?

To: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-users] How to get rid of IO-Bottleneck?
From: "L.M.J" <linuxmasterjedi@xxxxxxx>
Date: Fri, 3 Apr 2009 09:02:07 +0200
Delivery-date: Fri, 03 Apr 2009 00:03:08 -0700
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <49D5ACF9.70305@xxxxxxxxxxxxxxxxx>
List-help: <mailto:xen-users-request@lists.xensource.com?subject=help>
List-id: Xen user discussion <xen-users.lists.xensource.com>
List-post: <mailto:xen-users@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=unsubscribe>
References: <49D5ACF9.70305@xxxxxxxxxxxxxxxxx>
Sender: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
Le Fri, 03 Apr 2009 08:30:17 +0200,
Fabian Zimmermann <xen-users@xxxxxxxxxxxxxxxxx> a écrit :

> Hi,
> I'm running a Xen (Debian Lenny / 3.3.1 / / 2x1T SATA RAID1 / 
> 8G RAM)-Server with ~8-12 VMs (LVM based - NOOP shedule). Each VM is 
> running Postgre, Tomcat, Apache. If we do our daily testings, the 
> disk-io-performance  is terrible.
> As far as I understood reads are cached within the VMs, so this should 
> not be the problem. I think the high number of writes is killing the 
> hdd. Is there a way to do something like "async write", or what else 
> could I do to speedup the VMs even if there is a lot of disk-io?

 Hi Fabian,

   You could start to investigate with the "top" command and "iostat". Iostat 
will show you the actual disk 
throughput. If you near your disk maximal throughput speed, then, you can 
expect more from SATA HDs on
RAID1. Using top on each host, pay attention to the  "%wa", it's the percentage 
of wait, meaning the system
wait too much the storage. On a perfect system, "%wa" should stay down to 0.

   First of all,  I will buy another SATA disk to setup a RAID5 instead of 
RAID1, this will improve (a
little bit) I/O performance or better, RAID0+1. Do you use RAID Software 
between ? Maybe you could increase
database memory to decrease read I/O too. Finally, if you have some extra $, 
try to get a SAS RAID Controller
card such as [1] and a couple of SAS HDs (I would say at least 4) to setup a 
fast RAID5 storage group.

  Good luck,

[1] : 

Attachment: signature.asc
Description: PGP signature

Xen-users mailing list