Hi folks, I did this for about 6 months to do evaluations of Exchange 2010 and Zimbra.
My cluster had two VM hosts, each with 6 nics (2 onboard used for heartbeat, and an an in Intel PCIe quad port). I defined a LAN (vswitch) internal to the cluster only for traffic between all the VM's and the Lan side of the pfsense box. I also added one port from each of the VM hosts and connected to an external switch VLAN which was then directly connected to the internet. DRS and HA worked flawlessly. This worked exceptionally well for the pfsense box. The VM hosts were dual processor dual core P4 Xeon's at 3.0Ghz. The internet connection was 100Mbit and I was easily able to get 80+Mbit across it. CPU use on the VM was never more than 20% of the single vCPU I assigned to it. In the 6 months we had it running it never burped once. It performed exactly like a hardware box. I did not install the VMware tools on pfsense. I would not recommend this for a production scenario though, there are too many unknowns about the footprint that vmware might expose. Especially seeing any only computer will run pfsense very well if all you need is basic routing and NAT'ing. This was on VMware ESXi 4.0 hosts, with a single vSphere manager. We are currently playing with vyatta to do some really neat routing simulations for our larger network which is all cisco at the routing layer. We have several VRF's defined in our cisco's and have been playing with the open source patches to add this to the vyatta project that have not yet been integrated. For us, if we can prove this is stable in vmware, we will consider moving to hardware vyatta boxen. Good luck! Tim
