Diagnosing a bottleneck in a colocation channel?

Question

linux

There are several servers on linux, on colocation in the data center, they stand side by side, among them there is one, when it was plugged into another port in the cisco, it turned out, despite the fact that the link on the port to the cisco is defined as 1Gbit, and Here is the speed of return-reception of the order of 200 kbps.
“On the other side”, the sisamdins convinced that everything is fine on their side, there are no restrictions and shaping and there can be no, especially since neighboring servers give out the declared speed.
What tools can be used to diagnose the channel and identify where the plug is?
how to make diagnostics for a network card, if we assume that it is she who fails?
Physically, of course, there is no access to the car.
Network Ethernet controller: Intel Corporation 82574L Gigabit Network Connection

Answer 1

iperf is a good testing utility
Play around with the number of threads.
Just recently I solved such a question, scored a little exp.
Write in a personal if anything.