On Sun, Sep 19, 2010 at 12:58 PM, ma qiang <maqiang1984@xxxxxxxxx> wrote:
> Thanks for your replies again.
> I have a HVM domU already running. So I have
> tried http://pastebin.com/fb6fe631, and reboot domU successfully. But when I
> run iozone in this domU, I found the performance is worse.
This is a different issue then? Previously you only mention network
throughput problem.
> So I have no idea
> why. I will test the network IO later.
First, make sure that the PV drivers is indeed active, as shown from that link.
# ls -la /sys/class/net/eth0/device
lrwxrwxrwx 1 root root 0 Apr 25 11:50 /sys/class/net/eth0/device ->
../../../devices/xen/vif-0
# ls -la /sys/block/hda/device
lrwxrwxrwx 1 root root 0 Apr 25 11:50 /sys/block/hda/device ->
../../devices/xen/vbd-768
Obviously adjust the device names according to your setup (e.g. you
might have hda1, sda1, or xvda instead of hda). If the output is not
similar, then something's wrong.
After that, check your domU setup. IIRC it's quite "normal" to have
higher performance on domU compared to dom0 when you're using
file-backed domUs (i.e. domU's disk is located as a file on dom0).
This is due to caching effect, and does not show real performance. You
should either use LVM (or partition, or some other block-device), or
use "tap:aio:/" instead of "file:/" on domU config file.
Next, try the same benchmark on dom0. It's normal to have
dom0 throughput >= PV domU throughput > HVM with PV drivers > HVM
without PV drivers.
--
Fajar
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home