Just for a testing purpose I used Xen almost a year to consolidate my dedicated server, initially dual core and now quad core. First I used Xen VM migration feature which was allowed me to have a separate solution on a server to use DEV. After switching to Core2Duo, I took an advantage of virtualization for consolidation and disconnected the the four large machines that are now just more than a small NAS.
Recently, I decided to upgrade to a Core2Quad before the technology changed. Even though the principle defined for motherboard / cpu to 45nm demands high end servers.
During this change of hardware, I didn’t really expect that at one stage I will have to abandon Xen. Actually, while installing an OpenSuse 11.3, I found myself on a completely unstable system under Xen ( without VM ) which crashed when tens or hundreds of data have been transformed over the network. So finally I opted a more revolutionary solution – qemu-kvm hypervisor, based on its classic nature.
The migration wasn’t too difficult, as Xen & KVM work with a logical drive in a file. So I’ve had to modify the kernel to include a core as a “default” to reboot the system under qemu. In addition, using YaST tools to manage these Vms were simple and easy to use to create an example of VM description in XML format.
Since all of my VM work properly, what I’ve noticed, the configuration options in KVM are somewhat richer than that of Xen. By the way, all these tests tell me that a kernel module should be addressed in KVM, because Xen already shows a certain lack of stability and its implementation is also more complex than KVM.