On Fri, Jun 27, 2008 at 05:56:37PM -0400, Paul Wouters wrote: > Although I was very happy with xen, it seems xen is losing out and kvm > is the way to go. I had two migration questions that I couldn't find > answers for. > > 1) Does kvm also run with para_virt (eg without VT capable cpu, but also > just to be slightly more effecient). virt-manager only gave me a qemu > option on my old amd64 cpu. KVM requires a hardware virt-capable CPU. These are quite common, (if your CPU was made within the last two years or so) but note there are at least two major 'generations' of virt technology in CPUs. In particular very new CPUs from both Intel and AMD support some form of virtualized page tables, which can be very beneficial. Intel describes them here: http://www.intel.com/technology/itj/2006/v10i3/1-hardware/8-virtualization-future.htm [Question: does EPT / NPT only affect 32 bit?] The second area where Xen had a traditional advantage was in paravirtualizing disk, network and a few other devices. Linux now supports something called virtio, a standard for virtualized disk and network devices, and KVM (and others) support this. Here's some more info about virtio: http://lwn.net/Articles/239238/ As for your problem with virt-manager, can you please tell us what it says in /root/.virt-manager.log and the output of 'virsh -c qemu:///system capabilities' command, and the contents of /proc/cpuinfo. > 2) I boot all my xens using kernel+initrd outside a rootfs image, instead > of a bootable virtual disk. Is that possible with kvm, or will I be > stuck creating bootable disk images? KVM has some command line options to do this, namely -kernel and -initrd. Rich. -- Fedora-xen mailing list Fedora-xen@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-xen
