I'm planning to do a shoot-off of Windows Server 2008 Hyper-V vs VMWare ESX 3.5 on the same hardware to see how Microsoft's offering compares to ESX3i. I have my own opinion, but am willing to be proved wrong one way or the other.
To keep things as close to the same as possible, all VM's will be stored on the host's local disk, all VM's will be running together, and benchmarks will be between VM's on host, from VM's to an iSCSI target that will get rebuilt between tests, and from VM's to a Linksys gigabit NAS device.
VM's to be loaded into the hosts will be (hopefully) as follows;
1x SBS 2003 R2, 2GB vRAM, 100GB vdisk
1x Win2K3 R2, Terminal Server, 2GB vRAM, 20GB vdisk
1x SME Server 7.3 (for proxy), 512MB vRAM, 8GB vdisk
1x YAFPC PDF creator server, 256mb vRAM, 1GB vdisk
1x Symantec Brightmail v7 appliance, 1GB vRAM, 30GB vdisk
This will be set up similar to an environment I've just deployed for a customer on 'real' hardware hosting 15 users.
Haven't quite come up with the benchmarks as yet, maybe loading up Exchange, simulating Terminal Server users, file copies, simulating lots of inbound email. Something like that.
Each VM will be migrated P2V via ShadowProtect images - that way when they hit the host they'll be exactly the same.
I'll be monitoring stuff like host processor utilisation, disk throughput, network throughput, and from the VM's themselves how responsive they feel when the system is busy.
This should be fun.
Other related posts:
Life in the fast lane, not!
Why I love Blackberry Enterprise Server
Poor Aussie surfing for knockers at home - how to work around this yourself