Name Resolution Problems in Windows Domain

We've just replaced 20 Dell desktops running Windows XPsp2 with 20 iMacs running the latest 4.0 build of Parallels. Some of these machines are running Windows XPsp2 as the guest OS, others are running Windows 7. All VM's have been joined to our Windows domain. The iMacs are running OS X 10.6.1.

We have two subnets: 192.168.0.xx and 192.168.10.xx. We use DHCP to allocate addresses in both subnets and that works fine. We have a requirement to run the VM's in Bridged Networking mode so the workstations can see each other, our VOIP phone system, and our IP printers. Our domain controllers reside in the 192.168.0.xx domain. We've noticed that workstations in that subnet can run in Bridged Networking mode with no problems, but the workstations in the other subnet have almost constant name resolution problems resulting in dropped and reset network connections with our servers, the Internet, etc. If we switch those workstations to Shared Networking mode, the name resolution problems go away, but now they can't access our IP printers or shares on other workstations, and they can't receive notifications from our voicemail system. We really need Bridged Networking to work.

We kept 2 of our Dell computers and have verified that they can run in the 192.168.10.xx subnet with no problems. We also know that the Macs themselves operate just fine in both subnets. BTW, none of this is wireless -- all connections are wired Ethernet connections, and all airport adapters are turned off.

We're stumped. Any ideas would be welcome. Thanks.

 

It would be interesting to collect network traffic dump (for example using WireShark) in some of VM with problem (while it have problem). And also look at the ipconfig /all on the same machine - it may be something really simple - like two DNS-servers one of which is not functioning

 

one more thing to look is to start /Application/Utilities/Terminal at Mac side and execute command "prl_perf_ctl -a | grep nic".

Parallels does statistics of received/sent network packets and errors encountered.
pkt_dropped_in, pkt_err_in, pkt_err_out will definitely point to some problem

Another interesting counter (it is not an error, but it is interesting) is send_buffer_notifications. It should not grow too fast and likely should be zero unless VM sends packets at high rate