This simple experiment was done in Mininet using these inigo test scripts. Two 30 second iperf flows originating from different hosts, one arriving 10 seconds before the other, both heading to the same server host. You can see that TCP Inigo keeps the bottleneck queue on the switch small while achieving good and fair utilization.
This Mininet setup uses 700Mbps links and a 0.5ms RTT, and ECN is only enabled for experiments including "+ecn" in their name. So, for instance, a DCTCP experiment will act like TCP Reno unless it has "+ecn" in its name. Graphs are in *zoo* directories.
In fact, Inigo behaves similarly to a TCP that has ECN information, even though it is only using RTTs.
