Talked to a rep today at Virtual Computing.
There are some "soft gotchas" with the bare metal VM concept. ( A soft gotcha is one that if you gave it a second thought or two, you would've known about it before asking the question.)
The obvious one is "How do you deal with all the freaking hardware drivers?" You limit the hardware selections supported.
1) The product is tailored for specific sets of hardware from one of the big three OEMs- DELL, HP and Lenovo or full Intel chipset boards. No AMD or other second source x86 need apply. A Pentium 4 isn't going to hack it either. Systems supported are high performance desktops and laptops.
2) It requires an Intel VT equipped CPU. That limits it to the Core 2 Duo or Quad CPUs. (I knew that was going to be a possibility.) AMD multi-core compatibility is still being worked on.
3) Video adapters are limited to Nvidia and Intel chip sets. No ATI video boards supported at this time. Well I didn't think that was going to happen with AMD owning ATI.
4) The software works with a client and server model. The client connects to the server that delivers the VMs to the hypervisor client. The client is on the Intel VT
5) The Server application has to be running on Windows Server 2008 running on just about anything, no dual or quad CPU necessary. No plans at present for a Linux server capability.
6) Management module resides on the server. Features in the VM and hypervisor can be tailored like levels of security or access to the hardware based on user groups or logon. Active Directory was very faintly hinted at. Each VM can be limited in access to the peripheral hardware on the client machine.
7) VMs are pre-designed or can be client designed as "generics" or as custom as wanted. All VMs can have the same or differing levels of access on each client.
8) "Restore points" or backups of the VM as modified by the client-user can be scheduled or initiated by the user.
9) VM images can be set as permanent safety images with incremental updates to operate as a fallback to wipe out malware in the VM.
10) New software release is planned this coming Monday or Tuesday.
Still got some growth to do but it looks really interesting for a lot of reasons.
The biggest is that if its done right, it eliminates the double-API conversions that have to happen between the OS in the VM and the OS on the host hardware. That tighter binding to the hardware should offer a really nice speed improvement.
It might spur some activity by others in the VM game to bring out their products. My experience with VMware on XP Pro was not a bad event but it was still slow on certain software applications.
In many ways, a stripped down Linux works fairly closely to the hardware level of this no-OS hypervisor. Maybe a group of Linux driver and kernel programmers are working on something like this?