One big distinction you might want to make when it comes to virtualisation is "para-virtualisation" versus "full-virtualisation".

The difference is that a para-virtualised system is aware that it is virtual. This means that it can forego some tasks a non-virtualised system would have to do:

i.e. a disk driver needs to check if the disk is still accessible, may have to cover for a block becoming bad, etc.. A fully virtualised disk driver thinks that the (virtual) disk it is presented is in fact a real disk and treats it accordingly. It will do all these checks i talked before. A para-virtualised disk driver would "know" that what seems to be a disk is in fact a file on some real disk and forego all these checks because the disk driver of the host system would carry them out anyway. This will make the para-virtualised driver faster and less resource-consuming. On the other hand it would mean that you have a system with a different driver and you can't simply copy your physical system to a virtual environment (or the other direction) without changing some things.

I haven't worked with OpenVZ for a long time, so i cannot say anything about newer versions. OpenVZ is (was?) such a para-virtualised environment which offers performance benefits compared to fully virtualised environments like VirtualBox, etc.. But basically it is (or was?) a very comfortable chroot-environment and the performance gain meant on the other hand, that you had only one kernel really running. This meant that you were not completely independent in chosing the system you wanted to run virtualised. If it is now possible to mix Windows- and Linux-systems on the same hardware node IMHO this means that they have switched to full virtualisation and this perhaps means giving up some/all of the performance advantages the para-virtualised system once had.

But without more clearly defining what you want to do a serious suggestion is not possible, as already has been said times and again.

I hope this helps.

bakunin