Skip to main content

Concurrent call

11 replies [Last post]
arundeking
Offline
Joined: 2008-05-03

Dear All,
How many concurrency calls did mobicent server support.
Thanks

Reply viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
ian_c
Offline
Joined: 2007-12-09

Ok, now I better understand the aim of this performance test. Unfortunately I'm not so skilled to develop a benchmark on my own, but I see that somebody else here is considering the task. I look forward to see real life throughput performance figures for Mobicents!

arundeking
Offline
Joined: 2008-05-03

Hi Eduardomartins,

I have following queries, please clarify it.

1. I have a query on call, according to Sip the call duration means from the invitation request to the ACK req [b]or[/b] the total call duration(Invitaion req, ACK req, their voice transfer, finally Bye).

2. In Mobicent server at optimise condition how many con current call it can handle per second.

3. In Mobicent testing using SIPp, I have doubt in my flow.

a. For UAC,I have installed SIPp in a PC(192.168.1.120).
b. For UAS, I have installed SIPp in a PC(192.168.1.130).
c. Mobicent server is running in another PC(192.168.1.2).

UAC is run with following cmd : ./sipp -sf uac.xml 192.168.1.2 -s 1003 -i 192.168.1.120 -p 5062.
In uac.xml in "To Address" I have give UAS address(192.168.1.130).

UAS is run with default cmd ./sipp -sn uas.
(For UAS I couldnt execute custom scenario, if I execute with ./sipp -sf uas.xml, it displays "resolve ip address", please give me syntax to run custom scenario for UAS).

While executing in the UAS report "Unexpected Message" column its value is increasing and the Trying, Ringing, and ACK column have reached upto 295 and it stop, similarly in UAC also the Invite and Unexpected Message is increased but the ACK, Trying and Ringing are stopped at 295.

In the mobicent server the request has been transfered on both side.

Can you please brief me whether my flow is correct or any other steps to test the no. concurrent call handle.

Thanks

rayesh
Offline
Joined: 2006-05-03
eduardomartins
Offline
Joined: 2005-10-10

Actually that page needs some updates, a recent dual core laptop can do around 150 cps, a entry level server does 200 cps, but this really average data, you need to tune the container and see what you get.

http://groups.google.com/group/mobicents-public/web/mobicents-load-testing
http://groups.google.com/group/mobicents-public/web/mobicents-performanc...

ian_c
Offline
Joined: 2007-12-09

Hi, is it possible to understand exactly how the cps count is evaluated ? Do you have a published flow diagram to see what are the SIP packets exchanged and how a call is defined ?

Regards,
Ian

eduardomartins
Offline
Joined: 2005-10-10

CPS value is defined by SIPP, and you can find the messages exchanged at http://groups.google.com/group/mobicents-public/web/mobicents-load-testing

ian_c
Offline
Joined: 2007-12-09

Eduardo,
thanks for the answer. I see that the definition of CPS used in your test is a bit different from the one given in the sipstone test, and used as a reference for SIP based servers. This makes difficult to compare Mobicents SIP performance with other products, as the CPS given could be misleading.

Regards,
Ian

Message was edited by: ian_c

eduardomartins
Offline
Joined: 2005-10-10

Not getting, as far as I know SIPP is a test application that implements SIPStone, so the "New CPS" value should be the value you use in comparations. I'm missing something?

ian_c
Offline
Joined: 2007-12-09

It's just that for a production environment there should be some overhead due (for example) to multiple TRYING, call routing (for a proxy), CDR generation, rejected calls and tracking also partial timers such as RINGING, well you get the idea. A realistic benchmark like the one published in
http://www.transnexus.com/White%20Papers/2007-06-04_OpenSER-SER_Benchmar...

Although not a proxy server, my attention was raised because of the very good performance Mobicents has with respect to workhorses such as OpenSER and SER.

Message was edited by: ian_c

eduardomartins
Offline
Joined: 2005-10-10

Ours is just intended to be a SIP load test, to measure performance of the SIP Resource Adaptor, but you are free to contribute to such a kind of network load test. Anyway, in my humble opinion, that is not so much interesting for an SIP app server, which are our targets :-)

ian_c
Offline
Joined: 2007-12-09

Ok, now I better understand the aim of this performance test. Unfortunately I'm not so skilled to develop a benchmark on my own, but I see that somebody else here is considering the task. I look forward to see real life throughput performance figures for Mobicents!