We have 2 sites. Point A and point B. Point A has a 100M ethernet line on a 100M port and point B has a logically 20M line seating on a 100M port.

Telco providers use traffic shaping to logically set the bandwidth on their L2 device at their nearest office from point B. We have multiple circuits with this telco and they are using the same vlan within the telco NW side. From our side, we set our own vlan between our cisco devices at both point A and point B. All our traffic are transparent to the telco including our vlan tags. When we conducted our regular TCP test from server to server, it failed. We were not getting the expected bandwidth.From point A to point B, it was good but from point B to point A, we were getting only about 2M. The telco insist their comittment is up to L2, they have conducted throughput test sending 20M of frames across the wire and they were all good. But still, the issue remains on the same on our side, we've also tried traffic shaping on our side. From the telco perspective, they just prepare the road, and they let us drive whatever car we want.

 Any ideas or suggestions what could be causing this or has anyone experienced this kind of issue before?We use iperf for testing.

Thanks



Software/Hardware used:
Cisco 2980 at point A, cisco 2800 at point B