WARNING - OLD ARCHIVES

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/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-users

Re: [Re: [Xen-users] Slow domU network

To: aschiefer@xxxxxx
Subject: Re: [Re: [Xen-users] Slow domU network
From: Nivedita Singhvi <niv@xxxxxxxxxx>
Date: Sun, 19 Mar 2006 11:38:16 -0800
Cc: xen-users <xen-users@xxxxxxxxxxxxxxxxxxx>
Delivery-date: Sun, 19 Mar 2006 19:39:47 +0000
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <1141983797.20734.21.camel@stella>
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/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=unsubscribe>
References: <1141920017.19645.61.camel@stella> <1141921563.19645.64.camel@stella> <20060309183844.GA10563@xxxxxxxxxxxxxxx> <1141983797.20734.21.camel@stella>
Sender: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
User-agent: Mozilla Thunderbird 1.0.2-6 (X11/20050513)
Artur Schiefer wrote:
hello,

somehow the linux network core tries to adjust itself to its
environment. as i recently learned theese parameters are optimized for
100 mbit networking. from another poster i got the info that you have to
set the following parameters in your "/etc/sysctl.conf" to achieve gbit
speed:
# increase TCP maximum buffer size
net.core.rmem_max=16777216
net.core.wmem_max=16777216

# increase Linux autotuning TCP buffer limits
# min, default, and maximum number of bytes to use
net.ipv4.tcp_rmem=4096 87380 16777216
net.ipv4.tcp_wmem=4096 65536 16777216

I would highly recommend also increasing the default
size that the buffer is. The TCP buffer increase relies
on learning from the behaviour of the traffic - but
a large number of connections do not last very long.
If you have a moderate sized or smallish file to send,
you don't catch up fast enough.

Try:

net.ipv4.tcp_rmem = 1398080
net.ipv4.tcp_wmem = 1398080

thanks,
Nivedita

BUT THIS ONLY SOLVES THE PROBLEM FOR WITHIN DOMU COMMUNCATION (i.e.
sender and receiver are the SAME virtual maschine, which should be a
theoretical problem in most cases). I tuned domU and dom0 this way
problem between dom0 and domU still exists. i also played with tcp
window sizes (-w option in iperf) from 64kb to 2MB but the performance
differences are negligable.


Am Donnerstag, den 09.03.2006, 20:38 +0200 schrieb Pasi Kärkkäinen:

On Thu, Mar 09, 2006 at 05:26:03PM +0100, Artur Schiefer wrote:

ok, solved the problem for within domu communcation, was an autotuning
effect.


Autotuning effect? Please be more specific because many of the xen users
seem to have problems with network performance..

-- Pasi
^
                               .     .
                                Linux
                             /    -    \
                            Choice.of.the
                          .Next.Generation.

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users

<Prev in Thread] Current Thread [Next in Thread>