[Qemu-discuss] high load when rsyncing data

[Blaster]

Hello,

I am using Ovirt 3.3.2 on Fedora 19, latest updates.  The guest is Fedora 20, 
latest updates.
Hardware is an i7 950 with 24GB of RAM.

The guest has 3 disks attached, a 100G QEMU boot disk, and 2 3TB Toshiba SATA 
as Direct LUNs using VIRTIO SCSI.

The system has 2 physical network interfaces.  1 connected to just he 
hypervisor and 1 configured as an oVirt bridge.

When I run an rsync inside the guest form one disk to another, the load average 
on the Hypervisor goes from 1 to 30 in a matter of a few minutes.

qemu-system-x86 is the top load consumer.

Is this normal?  This seems like awfully high CPU load.

The real problem comes in when I try and copy data into the hypervisor using 
the hypervisor interface.  IP traffic on the oVirt bridge interface (a separate 
physical interface and bridge) comes to a complete standstill until I stop the 
transfer.

These are normal tasks I perform quite regularly on physical hardware.  I 
expect some overhead for virtualization, but these tasks are quite trivial for 
a system this big, I wouldn’t expect resource exhaustion like this.

How do I go about tracking down what’s going on?