[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] Xen network tune for fileserver in DomU



Alexandr,

Using Xen for a fileserver doesn't make much sense. It's an inefficient use of resources.

Peter

Sent from my iPhone

On Jun 25, 2009, at 10:47 AM, "Alexandr R. Ogurtzoff" <iscander@xxxxxxxxx > wrote:

Hello, gentlemen
I need wise advice about how to tune  network for performance usage
file server under domU
I use bridged network.
On dom0 network settings are:
auto eth0
iface eth0 inet static
address 192.168.1.100
netmask 255.255.255.0
network 192.168.1.0
broadcast 192.168.1.255
up /etc/xen/scripts/network-bridge start bridge=br0 netdev=eth0
up ifconfig br0 up
#up route add default gw 192.168.1.1
gateway 192.168.1.1
dns-nameserver 192.168.1.1

$ brctl show
bridge name bridge id STP enabled interfaces
br0 8000.000c76ba7b5e no peth0
vif1.0
vif2.0

I use two domU instances:
app 192.168.1.110
file 192.168.1.140

... and GlusterFS on file to export filesystem.

then test it if bonnie I got next results:

Version 1.03c ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP app.cluster.xe 128M 5623 17 16612 5 1818 3 15254 46 26112 9 123.5 0 file.cluster.x 128M 9854 29 21547 4 3515 3 16907 44 30468 7 188.1 1 iscander 128M 11308 37 11344 1 555 0 10589 33 11306 1 70.1 0 file.cluster.x 128M 22408 74 23801 11 14326 4 18696 43 48676 4 545.8 1 ------Sequential Create------ --------Random Create-------- -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files:max:min /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP app.cluster.xen 16 242 4 1106 5 300 4 242 4 1144 6 288 4 file.cluster.xen 16 873 8 4112 11 1009 6 792 8 4351 10 1165 8 iscander 16 565 2 2317 2 764 1 575 2 2546 2 755 1 file.cluster.xen 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ ++ + +++++ +++



first one then run on app instance with /data mount from "file" by glusterfs second then test with /import mount from "file" by itself by glusterfs
third line is then mount /data to my host in the same LAN (100Mbit)
last is then test _local_ /data on "file" (ext3) localy

I surprised why my external host works faster with mounted share then
internals - I mean located on the same physical node VMs?
How can I tune network stack for Xen and if it really Xen's network
implemenation penalty ?

/etc/xen/app.cluster.xen
# this is autogenerated by Ganeti, please do not edit
#
kernel = '/boot/vmlinuz-2.6-xenU'
ramdisk = '/boot/initrd-2.6-xenU'
memory = 256
vcpus = 1
name = 'app.cluster.xen'
vif = ['mac=aa:00:00:3e:98:e4, bridge=br0']
disk = ['phy:/var/run/ganeti/instance-disks/app.cluster.xen:0,sda,w']
root = '/dev/xvda1'
on_poweroff = 'destroy'
on_reboot = 'restart'
on_crash = 'restart'
extra = 'console=hvc0'

/etc/xen/file.cluster.xen
# this is autogenerated by Ganeti, please do not edit
#
kernel = '/boot/vmlinuz-2.6-xenU'
ramdisk = '/boot/initrd-2.6-xenU'
memory = 128
vcpus = 1
name = 'file.cluster.xen'
vif = ['mac=aa:00:00:d5:0f:6d, bridge=br0, ip=192.168.1.140']
disk = ['phy:/var/run/ganeti/instance-disks/file.cluster.xen: 0,sda,w','phy:/var/run/ganeti/instance-disks/file.cluster.xen: 1,sdb,w','phy:/var/run/ganeti/instance-disks/file.cluster.xen: 2,sdc,w']
root = '/dev/xvda1'
on_poweroff = 'destroy'
on_reboot = 'restart'
on_crash = 'restart'
extra = 'console=hvc0'
--
My best wishes.
Alexandr Ogurtsov.

Linux is very friendly it is just picky who its friends are

_______________________________________________
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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.