Coolness !

[Michael Welzl < >]

Hey,

I just wanted to share coolness   :)    two pictures that amaze me - maybe it’s just me …

I did some tests today to see if ns-3 is getting stable for real-life tests in the testbed. The attached images show 3 flows originating from ns-3, which are joined by 3 iperf flows (from Linux hosts) after around 20 seconds. One plot is a throughput plot, measured on the receiver side per port number, from the TEACUP testbed. The other plot shows cwnd values - the cwnd values of the iperf flows were obtained with ttprobe, whereas the three ns-3 flows are the normal cwnd ouput files from ns-3  !    :-)

(to be clear, for ns-3, it’s only one instance of an application running - just an ns-3 instance generating 3 senders)

Harder to see in the cwnd plots, and not extremely surprising: iperf is a bit more efficient than ns-3. Oh well  :-)   I don’t think we’re really going to let ns-3 *compete* against iperf/Linux for our results - rather, we’ll test each in isolation, or we’re probably measuring the efficiency of the ns-3 simulator more than anything else…

- but this seems to work really nicely!

Cheers,

Michael