Hi,
I'm doing the performance test with kamailio + RTPProxy, but I would
like to get the real calls count that the rtpproxy is serving. I don't
want to use value that I get from sipp.
So is there any management tool for rtpproxy, or should I get it
somewhere in kamailio config ?
Thanks,
Mino
You can't get it from rtpproxy. You'd really have to use something like the
dialog or htable modules to keep call state and get that from Kamailio.
-- Alex
--
Sent from my Samsung mobile, and thus lacking in the refinement one might
expect from a proper keyboard.
Alex Balashov -
2012/9/13 Alex Balashov abalas...@evaristesys.com:
You can't get it from rtpproxy. You'd really have to use something like the
dialog or htable modules to keep call state and get that from Kamailio.
On the contrary it's possible (using raw UDP reads/writes):
work ~: echo h1u203u03 I\n | nc -w
Really? Interesting, I had no idea. I thought the rtpproxy control protocol was
binary and did not lend itself easily to interaction in this manner. Thanks for
the tip.
-- Alex
--
Sent from my Samsung mobile, and thus lacking in the refinement one might
expect from a proper keyboard.
Ok, I'm tagging dialogs with dlg_manage(), but even if the call ends,
it still keeps info about this dialog in list kamctl fifo dlg_list.
Should I somehow close the dialog when the BYE transaction is ended ?
Peter: Thanks for the tip! Really interesting. But I do not
understand, why also this
If the dialog module is working correctly, and the SIP flows are
standards-compliant, the dialog module should automatically track all
subsequent state changes and remove calls from tracking after a BYE is
processed.
-- Alex
--
Sent from my Samsung mobile, and thus lacking in the
2012/9/13 Alex Balashov abalas...@evaristesys.com:
Really? Interesting, I had no idea. I thought the rtpproxy control protocol
was binary and did not lend itself easily to interaction in this manner.
Yep, this command protocol allows us to do tricks like this which is
quite helpful for
Thanks for this info.
Netcat is a really underappreciated tool. I can't count how many times I've
unnecessarily written utilities to send stuff on UDP sockets, forgetting that
nc can do it.
-- Alex
--
Sent from my Samsung mobile, and thus lacking in the refinement one might
expect from
I'm using rtpproxy_manage, so I assume unforce_rtp is not needed.
On Thu, Sep 13, 2012 at 4:10 PM, Peter Lemenkov lemen...@gmail.com wrote:
2012/9/13 Mino Haluz mino.ha...@gmail.com:
Peter: Thanks for the tip! Really interesting. But I do not
understand, why also this list contains the calls
Correct, but you still need to call rtpproxy_manage() on receipt of a BYE or
CANCEL. It'll just figure out what to do on its own.
None of this has to do with dialog state, though. Just rtpproxy control.
-- Alex
--
Sent from my Samsung mobile, and thus lacking in the refinement one might
Ok, so I put there unforce_rtp_proxy even though I'm using
rtpproxy_manage. The tip with nc now really shows the calls count.
But the dialog count is still higher and higher, so I have bug
somewhere in the configuration. I'll check it.
On Thu, Sep 13, 2012 at 4:53 PM, Alex Balashov
According to this
(http://transnexus.com/index.php/performance-test-results-for-openser-and-rtpproxy)
For a server hosting both OpenSER and RTPproxy, each 1 GHz of CPU processing
capacity can manage a maximum of 325 simultaneous calls.
I have 2.4GHz for rtpproxy, but CPU/Mem/network is ok, so
Hello!
2012/9/13 Mino Haluz mino.ha...@gmail.com:
The results:
- rtpproxy calls count 280
- sipp calls count 2000
- iptraf on proxy 4.8MB/s
- G711a codec
So if my calculations are right (16kB/s per stream * 280 = 4.5MB/s),
rtpproxy calls count is really the right value. CPU usage is ok
Thank you, setting ulimits worked! And the performance is the same as
stated in the document I mentioned! :)
One more thing, I found these errors in syslog:
Sep 13 18:38:18 perftest kamailio[5268]: ERROR: core
[parser/sdp/sdp.c:211]: Invalid payload location
Sep 13 18:38:18 perftest
I'd be curious to know if your performance might actually be higher, owing to
improved CPUs since the study was done, hyperthreading, more frames per NIC
interrupt, etc.
-- Alex
--
Sent from my Samsung mobile, and thus lacking in the refinement one might
expect from a proper keyboard.
El Thu, 13 Sep 2012 19:39:43 +0200
Mino Haluz mino.ha...@gmail.com escribió:
CPU E5504 Xeon 4MB cache 2GHz
800 concurrent calls with 92% CPU usage
So 400-450 calls for 1GHz (with 99% CPU usage, in fact I cannot get it
to 100% dunno why)
They had 325 calls per 1GHz, so yes, there must be
Hello,
On 9/13/12 9:22 PM, Ricardo Martinez wrote:
Hello.
Could it be that this number maybe represent a failed connection to
the radius server?
I am not using radius myself, but I would try some generic troubleshooting:
- watch the network traffic between with the radius server to see if
Hi,
Is there a module that can be used to collect number of MESSAGEs sent
or received by a subscriber;
number of registrations by a subscriber? These will be per subscriber
as oppose to statistics collected for
entire system.
Thanks
Krish Kura
___
SIP
Jon Bonilla (Manwe) writes:
There's another rtpproxy replacement which is compatible with kamailio's
rtpproxy module (no configuration changes needed in kamailio)
This one is threaded, works in kernel space with userspace failover and shoud
improve rtpproxy's performance.
how does that
19 matches
Mail list logo
