[asterisk-users] Asterisk V/s FreeSwitch

Jared Geiger jared at compuwizz.net
Thu Feb 9 12:06:25 CST 2012


We have used in production Asterisk 1.4 to do 3,000 concurrent calls at
about 80 CPS without media going through the system. This is on a vmware
ESXi server. The server is a Dell R610 with 2 X5670 (6 cores each at 2.93
GHz so 12 physical, 24 logical cores). Each vm gets 2 cores and 2 GB of
RAM. We put 3 of these VMs on the physical server with no issues. The CPU
usage above 3,000 channels starts to get out of hand. These were all ulaw,
no transcoding using directmedia=yes and canreinvite=yes

We tested on Asterisk 1.6 with the same sort of configuration. Asterisk 1.6
could not get above 1000 reliably. Even though it was configured to not be
in the media path, it would allocate a timer for each call which would
saturate the iowait on the VM. I verified that the VM wasn't processing
media.

We did not test on 1.8. We kept 1.4 up to date as possible and have moved
to an OpenSIPS routing platform instead to handle higher calls per second.

Regards,
Jared Geiger

On Thu, Feb 9, 2012 at 11:06 AM, Stefan Schmidt <sst at sil.at> wrote:

> Am 09.02.12 16:45, schrieb Patrick Lists:
> > Iirc a long time ago there was a discussion about load testing by
> > playing MoH was not a realistic test. Something about all MoH music
> > getting streamed synchronized so basically Asterisk only has to stream
> > one file and sorta multiplex that single output to all the established
> > calls (legs).
>
> this load tests are mostly about sip signal handling and not so much
> about rtp streaming but this moh class which i use had 100 files and
> random set to yes, so its atleast not soo bad.
>
> > [snip]
> >
> >> btw my normal production machines which are just the same virtual
> >> machines like this test system. i also had 330 concurrent calls, some
> >> with transcoding, many database lookups, musiconhold, pickup ... and the
> >> sysload was around 1.0 ;)
> >
> > The difference (13500 with MoH versus 330 with a real dialplan) shows
> > that it makes sense to mimic your dialplan in your test scenario as much
> > as possible to see how far you can realistically push the box and still
> > keep things stable and sound quality good.
>
> This 330 concurrent calls was only the highest value which i had on a
> normal production system and its really hard to build a test setup which
> presents a system with 4000 sip peer doing some calls.
>
> but the sound quality was still good even with 10000 calls in my tests.
>
> > Regards,
> > Patrick
> >
>
> best regards
>
> stefan
>
>
> --
> _____________________________________________________________________
> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
> New to Asterisk? Join us for a live introductory webinar every Thurs:
>               http://www.asterisk.org/hello
>
> asterisk-users mailing list
> To UNSUBSCRIBE or update options visit:
>   http://lists.digium.com/mailman/listinfo/asterisk-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-users/attachments/20120209/72f8e185/attachment.htm>


More information about the asterisk-users mailing list