[OpenSIPS-Users] Opensips + rtpproxy issue

Maxim Sobolev sobomax at sippysoft.com
Wed May 6 00:22:33 EST 2020


Hi Miha, sorry to hear about your issues. In order to troubleshoot it
further could you please also provide rtpproxy package version as reported
by the system package manager (apt, rpm etc) if the software has been
installed via that channel or branch name if it's been built from sources?
Unfortunately version reporting of the --version command has been bit
crippled until recently, already improved in latest master and 2.1 I
believe.

In general performance under virtual environment has not been terrific, due
to some design choices made early in our work. Hovewer I believe it should
be much better in 2.0 and 2.1 vs. 1.x series. Some of it is inherently due
to VM scheduling jitter, some is because we are unwilling to put it into
unsafe domain (i.e. kernel mode). As a rule of thumb, you might expect 3-5x
drop in max pps until jitter becomes an issue as compared to running on
comparable bare metal. Spinning multiple instances might help to mitigate
some of it though, but it also depends on hypervisor version and even
particular CPU generation.

-Max

On Tue., May 5, 2020, 6:10 a.m. Miha via Users, <users at lists.opensips.org>
wrote:

> Hello
>
> we have virtualized opensips and rtpproxy running on the same server which
> is virtualized in vmware infrastructure. Servers are not old, also traffic
> is not so big (cca 50 simultaneous calls). when there is a peak cca 80
> simultaneous calls RTP starts to break.
>
> is there any special setting/flag to be set, so that I can optimze this?
> load on VM is very low.
>
> rtpproxy -version
> Basic version: 20040107
>
> Opensips is 2.1
>
>
> thank you for help.
> Miha
> _______________________________________________
> Users mailing list
> Users at lists.opensips.org
> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/users/attachments/20200505/81454eae/attachment-0001.html>


More information about the Users mailing list