2. The Crossed Streams Jeopardy

Warning

To perform the crossed streams tests we had to reuse the topology prepared for The Mesh of Death Adversity Test because we were running out of time.

Therefore this scenario inherits all the settings of the previous test but adds an additional stream of traffic to stress test the network.

Topology

Same as The Mesh of Death Adversity Test with a small variation:

  • there is a second stream of iperf traffic from node D to H

Test

Note

The test scripts are available on github:

The tests mainly consisted in measuring ping Round Trip Time and throughput from client to server while generating two streams of traffic:

  1. from the client connected to A to the server connected K
  2. from node D (cross client) to node H (cross server)

The measurements where collected from A.

2 different tests were performed:

  • 10mbit iperf streams: measure ping RTT while 2 simultaneous 10 Mbit/s UDP Iperf streams cross the network
  • 100mbit iperf streams: measure ping RTT while 2 simultaneous 100 Mbit/s UDP Iperf streams cross the network

Note

Results

Graphs and raw data are provided for each test.

Note

The graphs were generated with the following command (requires the R programming language):

R --vanilla --slave --args --out-type svg --separate-output --maxtime 300 --maxrtt 500 --width 9 --height 5.96 --palette "#FF0000 #005500 #0000FF #000000" results/ < generic.R

the script generic.R is available on github.

10mbit iperf streams

Measured RTT in ECDF graph:

(How to read: closer to left is better, learn more about how to read ECDF graphs)

Measured RTT in classic graph:

(How to read: lower is better)

Measured Bitrate (from client to server):

(How to read: higher is better)

Note

Raw data for this test is available on github.

100mbit iperf streams

Measured RTT in ECDF graph:

(How to read: closer to left is better, learn more about how to read ECDF graphs)

Measured RTT in classic graph:

(How to read: lower is better)

Measured Bitrate (from client to server):

(How to read: higher is better)

Note

Raw data for this test is available on github.

Article written by Federico Capoano, Matthieu Boutier.