The advantage of RHEV for Servers / RHEV for Desktops is the management tools which allow you to quickly spin up nodes. It will be open source at some future date (when they can replace the .net dependency with JBOSS, last I heard ~ RHEV version 3.0)
RHEVD combines a couple of basic technologies KVM/QEMU, SPICE, KSM, NFS, Heartbeat and more. If you avoid HA you could just whack a bunch of desktops on each KVM host. Using RDP or VNC instead of SPICE and LVM snapshots to speed up the deployment. As long as your windows desktops are essentially stateless (ie network home, same image and directory bound) your able to simply put a load balancer in front of the hosts.
Keep in mind your contention ratios, KSM helps you over commit ram (ie 20 1gb vms in 10gb of system ram) but 1.5 - 2x over commitment is really the limit before performance starts plummeting with windows vms.
Also keep in mind 20 vms accessing a single disk or even a simple RAID5 array will likely result in IO being your major slowdown. RHEVD uses NFS for storage, but this is typically backed by a huge number of disks on a SAN / NetApp filer type system.
Virtual Desktop Infrastructure is typically used to reduce management requirements. Pragmatically If your only targeting 30 machines, then I am not sure the development required makes it worthwhile over other solutions (such as terminal services), not to mention Microsoft licensing for Windows VDI can be sticky if your not on a larger licensing agreement (OEM licenses can't be used).
Don't get me wrong RHEVD / VDI is an awesome technology, It just isn't a silver bullet. It works well in many situations but is definitely more suited to the larger environments (1,000+ desktops).
RHCE (RHEL 5)
Former Red Hat Technical Support Engineer