What are the two devices you are running iperf on?

What is a NID?

Here are my thoughts for cracking this problem (divide and conquer):

1. Plug the two notebooks in point to point and verify testing environment is good.
2. Plug both notebooks into the first switch device, and verify if performance is correct across this one device.
3. If the above test is good move one testing notebook into the next switching device in the path.  Repeat until issue happens.

Try and get to the point where you have the minimum amount of active network kit between the two notebooks to cause the problem to happen (you may have to move the first notebook to get to this minimal problem issue).  Then tell me what each of these devices is and how it connects to the next device.

Any chance you could re-format that last message?  Very hard to read.

When you say same segment test, what exactly are the two test points?  4500 to affected site?

This is starting to smell like a WAN issue ...

So we can rule out the 4507 and the connection through to the server.

The performance issue is only happening on the WAN portion of this test now, correct?

It just has to be a WAN issue.

A horrible test you could do, which you wont want to do, but will prove it.  Plug sites 'A' and 'B' WAN tails directly into a notebook and run iperf.  This way there is none of your kit in the picture and only the WAN.

If this fails (as expected) it rules out everything except the WAN.  It also means the WAN provider can refute it.  There is nothing else it could be.