Microsoft does not recommend clustering VMs, but does support it and it does work very well in actual practice since the environment remains available when patching and doing other maintenance activities - software availability versus hardware availability. The same reason to use VMs. You can run DEV, TEST, and PROD environments all on the same piece of hardware, but reboot each environment independently because of the VM. I was simply pointing to scalability, but if you also want reliability, no reason you couldn't add another piece of identical hardware and cluster the host server two-way. Of course, when you're talking about datacenter server, you're also looking at things like hot add for processors and memory and using MIBs to monitor the hardware to keep it highly available. You get the same hardware availability as a mainframe because of the management interfaces. You also get a technician from the hardware vendor onsite within hours with replacement parts when necessary. You're also able to do things like memory and processor mirroring inside the server chassis.