I know this post was started some time ago but thought I'd contribute some info that I haven't seen in the other posts in case the original author is still interested. I've run SQL on VMWare and XenServer and have a sharepoint server running on HyperV now but not hitting it too hard yet. I think from the database performance perspective you're going to see comparable results from any of the three hypervisors, depending on proper configuration, for small and medium sized databases. Can't speak to VLDBs.
The bigger difference between the three products are the features and the hassle associated with them. I really didn't care for XenServer from an administration standpoint, but have the least amount of experience there. HyperV is also more of a hassle administratively compared to VMWare what with the virtual machine config files getting corrupted easily and the increased necessity of host server reboots due to the monthly patch Tuesday Windows updates. Microsoft has available a "hypervisor only" verison of HyperV and maybe that requires fewer reboots and uses less host resources than when installing Server 2008 w/ HyperV, but I have no hands on experience with that one.
As for high availability, VMWare provides hardware layer redundancy, not application layer redundancy. Same for HyperV. VMWare Enterprise includes HA (High Availability), vMotion, and FT (Fault Tolerance).
With an HA cluster two or more ESX boxes are clustered together and a Virtual Machine running on one ESX host in the cluster will get migrated and booted up on another ESX host in the cluster in the case that the host machine it is running on goes offline. Notice the Windows VM has to boot up and will likely say to you "why was I shut down improperly before?"
With vMotion you migrate the VM in a controlled fashion from one ESX box to another with no loss of user session. You use this for planned hardware maintenance and ESX updates.
With FT you have a VM running on one host or ESX box and a mirror of that VM running on another box. If the host goes offline, client sessions get immediately redirected to the mirror and vSphere starts building a new mirror on some other ESX box in the cluster. This is like a combination of HA and vMotion and in fact uses the same technology.
But the bad news is... what happens if the VM bluescreens due to a virus or some other problem? Or what if you need to install Windows updates and reboot the VM? What if you have to install a service pack on SQL Server? None of the above help in that situation.
Still reading? OK, now I'm getting to the clustering question. VMWare does support MSCS or Failover Clustering as MS started calling it in Server 2008. But there are a lot of caveats. There's a "supported configurations" sheet that VMWare has out for each version of vSphere and none of them support Microsoft clustering when the shared storage is accessed over iSCSI or NFS or FCoE- only Fibre Channel is supported. Also, when using Fibre Channel they don't support NPIVs, which is very useful in virtual environments. (It's possible vSphere 5 now does support NPIV, I haven't read the specs closely).
All that "does not support" stuff being said, it all works. I've setup Failover Clusters using both iSCSI storage and FC storage with NPIV. all in vSphere 4/4.1. I'm sure VMWare would just say to you, if you called for support, that they can give "Best Effort" support only.
I know that was more like an article than a forum reply. In fact, I think maybe I have written about this elsewhere. But oh well, hopefully it is useful information.