Where are those outputs from? Maybe the authors replaced fmod/sec by flows/sec.
In practice, when the switch receives a reply (flow-mod), it'll forward the packet that caused it to send the packet in to the controller. Therefore it can be counted as a "flow". PUZZLE <puzzle.om...@gmail.com> escreveu no dia quinta, 26/05/2016 às 16:41: > Hi, > Thank you for the response, > The use of cbench is well defined in many documentations but I couldn't > find explanation about its given output results, > For example in these lines: > > *20:37:19.352 1 switches: flows/sec: 101527 total = 10.137119 per ms * > *20:37:29.458 1 switches: flows/sec: 95083 total = 9.502979 per ms * > *20:37:39.558 1 switches: flows/sec: 111017 total = 11.101199 per ms* > > *...RESULT: 1 switches 9 tests min/max/avg/stdev = > 547.22/11101.20/8529.18/3085.47 responses/s* > > it's giving flow/sec not fmod/sec, and then i find in previous work that > they are plotting "flow/sec" not "responses/sec min/max/avg/..", > > So i got lost in understanding how they get those metrics ?? > > If I am not wrong, and as i understood from Wcbench's plotter code > <https://github.com/dfarrell07/wcbench/blob/master/stats.py>, We only > plot the last line metrics "RESULT:", but they'r min/max.. > > > 2016-05-26 16:26 GMT+01:00 André Mantas <andremant...@gmail.com>: > >> I can try to explain what I understand: >> >> From the README page of cbench we get >> >> ./cbench -p 54321 >> cbench: controller benchmarking tool >> connecting to controller at localhost:54321 >> faking 16 switches :: 16 tests each; 1000 ms per test >> starting test with 0 ms delay after features_reply >> debugging info is off >> >> >> In this case cbench is running with default params (only port is set), >> which means it is in latency mode. In this mode each switch sends a >> PacketIn and waits for a reply before sending the next one. In throughput >> mode, switches send as many packets as possible at once (they don't wait >> for a reply). >> >> 16 switches: fmods/sec: 4661 4656 4655 4651 4649 4647 4645 4643 >> 4640 4636 4634 4631 4625 4621 4617 4608 total = 74.218852 per ms >> >> >> This line represents one loop execution of cbench. Each value is the >> number of flow-mods (and/or packet-outs after looking at the code) that >> each one of the 16 switches received received in one second. >> >> RESULT: 16 switches 16 tests min/max/avg/stdev = >> 74218.85/77850.14/76432.01/1214.77 responses/s >> >> >> In the end we get the total number of responses (min, max, avg and >> standard deviation) per second of all switches in all tests/loops. >> >> PUZZLE <puzzle.om...@gmail.com> escreveu no dia quinta, 26/05/2016 às >> 02:20: >> >>> Hi, >>> I am trying to plot the cbench results output but i still didn't quite >>> understood it especially due to the lack of documenttion, >>> >>> Would anyone please help me understand the numbers so i can can decide >>> what to plot? >>> >>> Thank you very much! >>> _______________________________________________ >>> openflow-discuss mailing list >>> openflow-discuss@lists.stanford.edu >>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss >>> >> >
_______________________________________________ openflow-discuss mailing list openflow-discuss@lists.stanford.edu https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
