Windows Terminal Server Load Balancing problem - FIXED (at last) I have deployed 2 Windows 2003 terminal servers running in unicast NLB mode, with 1 file server for the profiles etc. and am experiancing some performance issues. The TS boxes have a tendency to go slow then fast relating to network access. But they spend more time slow than fast. I suspect it is because it is trying to communicate via the NLB adapter not the management NIC like it sholud. The 2 TS boxes have 2 NIC's each (broadcom gigabit), one NIC is a managment NIC so it can talk to the other TS box, and one NIC is part of the NLB unicast team. As per the design requirements of Windows NLB The file server runs 2 NIC's in a load balanced/fault tolerance team (using HP NIC software). All the IP addresses of the servers and clients are in the same subnet (not how it should be but can't be changed at the moment). What is happening is that the NLB traffic tends to get broadcast all over the core switch cluster and out to the edge switches and routers. Performance can be very slow at times and lightning quick at some other times. the NLB IP is only set to load balance on port 3389 I dont think that he broadcast traffic is causing the problem from a network load perspective, but it might be in other ways. I have setup a few IIS NLB teams in an identical manner and have not run into this problem. Can anyone please help.. or report how they have setup a similar environment. This is acting rather weird.. im a linux engineer.. and have got a couple of windows engineers to come in and look and they can't work it out..