Any news?
A 12h16 Seg, 23 de Mai de 2016, André Mantas <andremant...@gmail.com>
escreveu:
> I've adapted cbench.py to handle OF1.4, can you take a look? Not sure
> about the match, I guess default values are ok?
>
> Then I ran both cbench versions, this time in *throughput* mode, with the
> respective cbench app running and the results are a lot closer now (full
> output attached).
>
> Original cbench with cbench.py running:
> RESULT: 16 switches 15 tests min/max/avg/stdev = 11959.97/12808.28/
> *12390.28*/233.99 responses/s
>
> Modified cbench with cbench_14.py running:
> RESULT: 16 switches 15 tests min/max/avg/stdev =
> 8763.60/9798.35/*9278.47*/291.91
> responses/s
>
> André Mantas <andremant...@gmail.com> escreveu no dia segunda, 23/05/2016
> às 11:24:
>
>> Thanks for your help and sorry for the delay.
>>
>> The -m cProfile worked. I've attached the results of the profile and the
>> output from both cbench versions. This time I tested with less switches and
>> less loops but the results are almost the same (simple_switch.py sends ~2x
>> more replies).
>>
>> Is it possible that simple_switch_14.py thinks that the switch has the
>> flow rules in its table? Because looking at the flows/sec for each switch
>> in cbench output, only the first loop has high values.
>> It might have to do with the features reply I send though.
>>
>> Victor Orlikowski <v...@duke.edu> escreveu no dia domingo, 22/05/2016 às
>> 00:49:
>>
>>> On May 19, 2016, at 6:08 PM, Victor J. Orlikowski <v...@duke.edu> wrote:
>>> >
>>> > A quick mailing list search suggests that there was a patch submitted,
>>> back in 2013, to have all of Ryu be able to be run under cProfile; it
>>> doesn't appear to have been committed.
>>> > It might make sense, though, to wrapper the appropriate portions of
>>> simple_switch.py and simple_switch_14.py in cProfile or hotshot, to see if
>>> there are significant differences in performance between the two (and - if
>>> so - where the time is being spent).
>>>
>>> A quick followup...
>>>
>>> If you can, it may make sense to run ryu-manager using an invocation
>>> similar to the following:
>>> python -m cProfile -s tottime ryu-manager ...
>>>
>>> It *should* be possible to profile your apps that way - but I have not
>>> actually *tried* it. ;)
>>>
>>> Best,
>>> Victor
>>> --
>>> Victor J. Orlikowski <> vjo@[cs.]duke.edu
>>>
>>>
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity
planning reports. https://ad.doubleclick.net/ddm/clk/305295220;132659582;e
_______________________________________________
Ryu-devel mailing list
Ryu-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ryu-devel
