Hi Tung
As you know Mediaproxy is written in Python, Python does not really support proper threading as it still has a global lock on the data dictionary.
This means that a SMP machine really does not help in fact because of the CPU switching it will probable perform to a lower level.
I would switch your machines make your test machine your production and then buy a second machine of the same spec and use the SRV record load balancing to give you both higher performance and redundancy.
All in all your better off having 3 or 4 machine than one machine in the long term any :)
Cheers Mike
Tung Tran wrote:
Hi all,
I've tested the performance of mediaproxy as below:
1/ PC1: Single P4, 2.4Ghz , 1GB memory, it can handle 420 sessions using rtpgenerator with g729. (testing server) 2/ PC2: Dual-xeon 3Ghz, 2GB memory, but it just can handle about 50 sessions when I use rtpgenerator with g729. (production server)
Should anybody can help me to customize my production server to handle more sessions.
Thanks and regards
Tung.
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Mike O'Connor wrote:
Hi Tung
As you know Mediaproxy is written in Python, Python does not really support proper threading as it still has a global lock on the data dictionary.
This means that a SMP machine really does not help in fact because of the CPU switching it will probable perform to a lower level.
Does this really cause a performance drop from 420 to 50 sessions?
regards klaus
I would switch your machines make your test machine your production and then buy a second machine of the same spec and use the SRV record load balancing to give you both higher performance and redundancy.
All in all your better off having 3 or 4 machine than one machine in the long term any :)
Cheers Mike
Tung Tran wrote:
Hi all,
I've tested the performance of mediaproxy as below:
1/ PC1: Single P4, 2.4Ghz , 1GB memory, it can handle 420 sessions using rtpgenerator with g729. (testing server) 2/ PC2: Dual-xeon 3Ghz, 2GB memory, but it just can handle about 50 sessions when I use rtpgenerator with g729. (production server)
Should anybody can help me to customize my production server to handle more sessions.
Thanks and regards
Tung.
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Hi Klaus
I've not actually tested Mediaproxy my self on SMP, but I have found this with other applications and Python.
Mike
Klaus Darilion wrote:
Mike O'Connor wrote:
Hi Tung
As you know Mediaproxy is written in Python, Python does not really support proper threading as it still has a global lock on the data dictionary.
This means that a SMP machine really does not help in fact because of the CPU switching it will probable perform to a lower level.
Does this really cause a performance drop from 420 to 50 sessions?
regards klaus
I would switch your machines make your test machine your production and then buy a second machine of the same spec and use the SRV record load balancing to give you both higher performance and redundancy.
All in all your better off having 3 or 4 machine than one machine in the long term any :)
Cheers Mike
Tung Tran wrote:
Hi all,
I've tested the performance of mediaproxy as below:
1/ PC1: Single P4, 2.4Ghz , 1GB memory, it can handle 420 sessions using rtpgenerator with g729. (testing server) 2/ PC2: Dual-xeon 3Ghz, 2GB memory, but it just can handle about 50 sessions when I use rtpgenerator with g729. (production server)
Should anybody can help me to customize my production server to handle more sessions.
Thanks and regards
Tung.
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
On 7/30/07, Tung Tran tr.tung@gmail.com wrote:
Hi all,
I just install another distributor linux: Red hat Enterprise Advance server 4, and got better result. The same server can handle about 400 sessions, (my first test was runing on CentOS 5) Yes Mike is right, SMP does not help to handle more session than single CPU server.
Hi Tran
You mean to say Media proxy with single processor
can handle 400 concurrent sessions, that very high compare to rtpproxy
can some one confirm, is the same case in production ?
ram
Hi,
I´ve seen rtpproxy with more then 400 concurrent session, without any problems.. so I know rtpproxy can do it :) (not sure about mediaproxy though)
- Atle * ram talk2ram@gmail.com [070730 10:15]:
On 7/30/07, Tung Tran tr.tung@gmail.com wrote:
Hi all,
I just install another distributor linux: Red hat Enterprise Advance server 4, and got better result. The same server can handle about 400 sessions, (my first test was runing on CentOS 5) Yes Mike is right, SMP does not help to handle more session than single CPU server.
Hi Tran
You mean to say Media proxy with single processor
can handle 400 concurrent sessions, that very high compare to rtpproxy
can some one confirm, is the same case in production ?
ram
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
On 7/30/07, Atle Samuelsen clona@cyberhouse.no wrote:
Hi,
I´ve seen rtpproxy with more then 400 concurrent session, without any problems.. so I know rtpproxy can do it :) (not sure about mediaproxy though)
400 sessions in Single server ?
what config
ram
* Tung Tran tr.tung@gmail.com [070730 12:18]: Hi Tung,
try sipp (sipp.sf.net) It has a rtppacket echo generator thing.. so, if you make a scenario and just call via your sip-server etc.. make sure the calls gets mediaproxied.. you can get rather good stats out from it. (I think)
-Atle