It helped me a lot (but did not really fix the problem) when I set my
disk queueing mode to deadline, from whatever it had been before. For
instance, on my domU:
echo deadline > /sys/block/xvda/queue/scheduler
The command is similar on the dom0, just use the appropriate block
device.
BUT, what *really* helped me out was to turn on write caching on my
SATA raid card. It helped such an absurd amount that I have to
suspect some corner case performance problem was being hit before,
but I don't know what it would have been.
On May 31, 2006, at 11:07 PM, Ashe Canvar wrote:
Hi all,
My dom0 and domU machines are showing terrible disk performance
degradation.
For instance my avg data read rate ( tested using hdparm -t /dev/hda)
is 55MB/s using the regular smp kernel but it falls to 2.6MB/s when I
boot into dom0. No other domains are runnign at this point. I have
looked for incrementing interrupts in ide0 , there are none. hdparm
shows that dma mode is enabled. dmesg shows that the same driver is
being used for both smp and dom0 cases ( Uniform Multi-Platform E-IDE
driver Revision: 7.00alpha2 ).
I have total of 2Gig RAM in the box and dom0 is restricted to 512M.
Any help appreciated !
Thanks,
-ashe
[root@medusa1 ~]# rpm -qa | grep -i xen
kernel-xenU-2.6.16-1.2080_FC5
kernel-xen0-2.6.16-1.2080_FC5
xen-3.0.1-4
--
Fedora-xen@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-xen
