[zeromq-dev] polling performances on windows

Francesco francesco.montorsi at gmail.com
Tue Nov 14 19:26:07 CET 2017


Hi Bruno,
I noticed your email and this reminds me of a performance issue with
polling that I hit recently with ZMQ.
Maybe you can be interested to the following excerpt:

-----
just to update on this topic, in case it's useful to others: it turned out
that the bottleneck [...] was more on the fact that I was calling
zmq_poll() before each zmq_msg_recv() (to implement a timeout on the
recv()).
The way I fixed the problem was to implement a slightly smarter polling
logic: I try to receive with ZMQ_DONTWAIT flag. Then if 0 messages are
received, next time a poll() operation with my desidered timeout will be
performed before attempting the recv(). If 1 message is received instead
then the next time the same recv() (always with ZMQ_DONTWAIT flag) will be
repeated.
-----

of this email thread:
https://lists.zeromq.org/pipermail/zeromq-dev/2017-October/031974.html
<https://lists.zeromq.org/pipermail/zeromq-dev/2017-October/031974.html>
This changed substantially the performances of my application. But perhaps
looking at your testing application, you are adding the zmq_poll()
operation on purpose, to test its impact so maybe what I wrote above does
not solve anything for you, not sure :)

HTH,
Francesco


2017-11-13 20:55 GMT+01:00 brunobodin . <brunobodin at gmail.com>:

> Hi all
>
> I ran a couple of test in order to evaluate the cost of polling (on
> windows). To do so, I added polling to the local_lat and local_thr tests.
>
> The code is here
>
> https://github.com/bbdb68/libzmq/tree/test_polling_cost
>
> and here is what I noticed :
> * since the fix about mempcy of FD_SET structure, the performances of
> local_thr are excellent,
> close to 1Gb/s ie the hardware thoughtput.
>
> * when I add the polling, the latency tests seems unaffected, while the
> thr test falls to 200Mb/s,
>   that is a 5x drop
>
> So here are my questions
>
> * is this way of testing polling meaningful ?
> * how do you explain the difference between latency and thoughput tests
> behaviour ?
> * what are the result on a linux box ?
>
> Thanks
>
> Bruno
>
>
>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail> Garanti
> sans virus. www.avast.com
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> <#m_-6006190082856671083_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> https://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20171114/be20f065/attachment.htm>


More information about the zeromq-dev mailing list