It is interesting, but ultimately I think the hardware solutions have more promise because they’re more general and require very little effort on anyone’s part to get the best performance.
Plus, the odds of MS letting someone modify the Windows source is… slim…
The big question is, when will the virtualization vendors be able to virtualize the video card, rather than emulate a mid-1990s, very basic card. Once Vista/Aero Glass hit, working in virtual machines is going to be a lot less enticing until they figure out the video issue.
I’ve had good luck with Linux VServer – multiple completely separate systems running on the same hardware, sharing the same kernel. Very low overhead. It’s best for webhost-type situations, where you have lots of somewhat-similar things running, but each system should be separately “jailed” away from the others.
It’s kind of infuriating when you’re searching for good heuristics for the A* algorithm, because searching for “a-star” gets you all sorts of results with “…and a star was…”
Searching for “A*” gets you everything that starts with A.
Take a look at Xen. It’s interesting because it offers better performance than traditional virtualisation systems by modifying both host OS and guest OS. It’s really interesting to learn about its optimisation techniques.