Even its a bit older,
1 i agree complete with olorin about upstream.
2 it can also be cause by - older kvm/libvirt on the host and new kernel on the guest (didnt even boot)
and of course vice versa
IO sheduler never use dealine, big disk operations lock everything up, a bit more inteligence is needed on the host.
true RHLE and fedore perform better regardnig kvm, for some reason debian in special but also ubuntu are vey slow with upgrades here.
they simply use to old versions, man even the driver cd for windows (kvm driver) is a simple link on deforas page, they could maintain a little better thats true
personally i used to use debian as host because its the fastest distribution and in my eyes also the most stable, but because of those update issues i went now to ubuntu.
iam really not afriend of ubuntu server. but of course i dont see fedora as a replacement its more an unstable testing ground, i dont want em in production
redhat well .. beside the tehnical aspekt, my problem is their constant widening of the renew stream.
serisously, its ok xxx costs a year, but only if i can be shure that this is stable for the next 10 years. so i know what iam dealing with the next 3-4 upgrades.
thing is you cant be shure with redhat, they come up with something new every version, pay per server, then pay per socket and per guest or take the big enterprise and so on
so honestly, as much i love kvm, since i cant see a way to adopt RHEL i either hope ubuntu/debian get their jobs better done
or i switch to vmware, which is in testing atm.... well ill give it another 8-12 months,