With the recent release of Fedora 17 comes a new KVM storage stack called virtio-scsi. Using this functionality is currently limited to command line operation. However there's already virtio-scsi code in upstream libvirt so virt-manager and other libvirt based tools will be supporting virtio-scsi soon.
One caveat about this new stack is that it does require guest drivers and Fedora 17 is the only guest so far supporting it. Virtio-scsi drivers will be available for linux guests from kernel version 3.4 so check your linux version if you'd like to try it. Guest drivers for windows however are not available yet.
You might be asking yourself why a new storage stack for KVM. The primary reasons the KVM developers came up with a new storage stack was more about the major shortcomings of the virtio-block stack. These shortcomings were listed in a mailing list discussion describing the virtio scsi design draft specifications and are shown below.
Limitations of virtio-block:
This doesn't mean that virtio-block will be going away though, it will still be useful for non-scsi block devices.
To verify that your version of qemu supports virtio-scsi devices run the following command on your linux host.
qemu-kvm -device ?
Piping this output to grep doesn't do any good so you'll have to scroll the long list of devices for the following output
name "virtio-scsi-pci", bus PCI, alias "virtio-scsi"
Once your output shows that device, your version of kvm supports virtio-scsi devices.
To get started download the recently released Fedora 17 iso to your linux kvm host. Next invoke kvm using a command line like the following where -cdrom argument points to your iso file.
qemu-kvm -drive if=none,id=hd,file=/path/to/fedora17.img \ -device virtio-scsi-pci,id=scsi \ --enable-kvm \ -device scsi-hd,drive=hd \ -cdrom /path/to/fedora17.iso
Go through the process of installing Fedora 17. When you're finished installing Fedora 17, reboot and your Fedora 17 guest with the following command
qemu-kvm -drive if=none,id=cd,file=/path/to/fedora17.iso \ -device scsi-cd,drive=cd,bootindex=0 \ -drive if=none,id=hd,file=/path/to/fedora17.img \ -device virtio-scsi-pci,id=scsi \ --enable-kvm \ -device scsi-hd,drive=hd
Virtual disks in your guest will be shown as /dev/sdX devices rather than /dev/vdX when using virtio-block.
It would be much nicer to test this with a management tool like virt-manager, proxmox or oVirt but new features are always available in command line and libvirt long before the management tools. Give it a try and please post any comments or questions you have.