100 parallel calls not possible (Performance problems)
Hello, we have a FreeSWITCH instance running on a IBM XSeries 32 bit server machine with a Windows 2003 operating system (RAM 6 GB). Our business logic is written in C# and integrated as a DLL module. Our module is originating the call via Native.Api DLL. With this solution we could make up to 30 parallel calls without any problems (calling a number, playing a wav file, hang up). If we increase the parallel call count to 100, we have problems with the sound (choppy) and the originate command returns with NO_ANSWER in many cases, but the callee has successfully answered (we have got many feedbacks, why we are calling but not speaking). In some documents and maillist postings we readed out, that freeswitch can handle more than 1000 sessions. Are this numbers only for linux machines? What are the limits of parallel calls with a Windows server and how we could increase the performance? Which environment is recommend for use FreeSWITCH with bulk calls? Does the .NET environment is blocking freeswitch while working? Best regards Ali
How quickly are you creating those one hundred calls? Pj On Mon, Jul 26, 2010 at 1:30 PM, Durmuş Ali Öztürk <ali.stgt@gmail.com>wrote:
Hello,
we have a FreeSWITCH instance running on a IBM XSeries 32 bit server machine with a Windows 2003 operating system (RAM 6 GB). Our business logic is written in C# and integrated as a DLL module. Our module is originating the call via Native.Api DLL. With this solution we could make up to 30 parallel calls without any problems (calling a number, playing a wav file, hang up).
If we increase the parallel call count to 100, we have problems with the sound (choppy) and the originate command returns with NO_ANSWER in many cases, but the callee has successfully answered (we have got many feedbacks, why we are calling but not speaking).
In some documents and maillist postings we readed out, that freeswitch can handle more than 1000 sessions. Are this numbers only for linux machines?
What are the limits of parallel calls with a Windows server and how we could increase the performance?
Which environment is recommend for use FreeSWITCH with bulk calls?
Does the .NET environment is blocking freeswitch while working?
Best regards Ali
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
The phone numbers are hold in a array, in a forEach loop the calls are submitted. Before calling, there is a wait time of 1 second. Also the calls are created very quickly. -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... Sent from the freeswitch-users mailing list archive at Nabble.com.
you may want to have a look at http://jira.freeswitch.org/browse/FSCORE-626 This has a compiler flag to improve timing on Windows 2003 and other older OS. -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... Sent from the freeswitch-users mailing list archive at Nabble.com.
The compiler flag is not found in my solution, should I get the latest source code? In case of a timing problem, why it does not occured while calling parallel 30 number? -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... Sent from the freeswitch-users mailing list archive at Nabble.com.
The compiler flag is not found in my solution, should I get the latest source code
That would seem like a reasonable thing to do.....
In case of a timing problem, why it does not occured while calling parallel 30 number?
In my (limited) experience there are two important things to control: 1) call per second 2) Total number of active sessions Calls per seconds appears (to me) to have a greater effect on the CPU. So at 1 call per second I would be looking at your CPU utilization and if it is maxing out then reducing that rate. My point being here that your box *may* be capable of handling 100 or 1000 sessions, if the cps is lowered. HTH On Tue, Jul 27, 2010 at 2:39 AM, Durmuş Ali Öztürk <ali.stgt@gmail.com>wrote:
The compiler flag is not found in my solution, should I get the latest source code?
In case of a timing problem, why it does not occured while calling parallel 30 number? -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... Sent from the freeswitch-users mailing list archive at Nabble.com.
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I will try to set up a fs system on CentOS, maybe the performance will be better. I am sorry but I dont understand the difference between cps and number of active session. If fs has 10 active sessions (with media), then the call per second should be 10 too?? 2010/7/28 Phillip Jones-2 [via freeswitch-users] < ml-node+5344088-1127633527-580578@n2.nabble.com<ml-node%2B5344088-1127633527-580578@n2.nabble.com>
The compiler flag is not found in my solution, should I get the latest source code
That would seem like a reasonable thing to do.....
In case of a timing problem, why it does not occured while calling parallel 30 number?
In my (limited) experience there are two important things to control:
1) call per second 2) Total number of active sessions
Calls per seconds appears (to me) to have a greater effect on the CPU. So at 1 call per second I would be looking at your CPU utilization and if it is maxing out then reducing that rate.
My point being here that your box *may* be capable of handling 100 or 1000 sessions, if the cps is lowered.
HTH
On Tue, Jul 27, 2010 at 2:39 AM, Durmuş Ali Öztürk <[hidden email]<http://user/SendEmail.jtp?type=node&node=5344088&i=0>
wrote:
The compiler flag is not found in my solution, should I get the latest source code?
In case of a timing problem, why it does not occured while calling parallel 30 number? -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html<http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html?by-user=t> Sent from the freeswitch-users mailing list archive at Nabble.com.
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=1>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org <http://www.freeswitch.org?by-user=t>
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=2> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org <http://www.freeswitch.org?by-user=t&by-user=t>
------------------------------ View message @ http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... To unsubscribe from Re: 100 parallel calls not possible (Performance problems), click here<http://freeswitch-users.2379917.n2.nabble.com/subscriptions/Unsubscribe.jtp?code=YWxpLnN0Z3RAZ21haWwuY29tfDUzNDA5OTZ8ODc3NDczNzM3>.
-- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... Sent from the freeswitch-users mailing list archive at Nabble.com.
CPS stands for Calls Per Second. Which basically means, the number of "new call attempts" that are hitting "every second". Suppose you have 1 call attempt coming in every second. After 30 seconds, out of the 30 attempts in total, only 10 were successfully connected. That would still make your CPS = 1. So, having 10 "active sessions" is not important. The CPU has the most work to do during call setup, as it has to process the logic of establishing the call. However, once the call is connected, there is very little "logical processing" for the CPU. Regards HASSAN On Wed, Jul 28, 2010 at 12:54, Durmuş Ali Öztürk <ali.stgt@gmail.com> wrote:
I will try to set up a fs system on CentOS, maybe the performance will be better.
I am sorry but I dont understand the difference between cps and number of active session.
If fs has 10 active sessions (with media), then the call per second should be 10 too??
2010/7/28 Phillip Jones-2 [via freeswitch-users] <[hidden email]<http://user/SendEmail.jtp?type=node&node=5345283&i=0>
The compiler flag is not found in my solution, should I get the latest source code
That would seem like a reasonable thing to do.....
In case of a timing problem, why it does not occured while calling parallel 30 number?
In my (limited) experience there are two important things to control:
1) call per second 2) Total number of active sessions
Calls per seconds appears (to me) to have a greater effect on the CPU. So at 1 call per second I would be looking at your CPU utilization and if it is maxing out then reducing that rate.
My point being here that your box *may* be capable of handling 100 or 1000 sessions, if the cps is lowered.
HTH
On Tue, Jul 27, 2010 at 2:39 AM, Durmuş Ali Öztürk <[hidden email]<http://user/SendEmail.jtp?type=node&node=5344088&i=0>
wrote:
The compiler flag is not found in my solution, should I get the latest source code?
In case of a timing problem, why it does not occured while calling parallel 30 number? -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html<http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html?by-user=t&by-user=t> Sent from the freeswitch-users mailing list archive at Nabble.com.
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=1>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org<http://www.freeswitch.org?by-user=t&by-user=t>
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=2> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org<http://www.freeswitch.org?by-user=t&by-user=t&by-user=t>
------------------------------ View message @ http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5344088.html<http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5344088.html?by-user=t> To unsubscribe from Re: 100 parallel calls not possible (Performance problems), click here.
------------------------------ View this message in context: Re: 100 parallel calls not possible (Performance problems)<http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5345283.html>
Sent from the freeswitch-users mailing list archive<http://freeswitch-users.2379917.n2.nabble.com/>at Nabble.com.
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Hi, have you checked your switch.conf.xml configuration for "sessions-per-second" value? <param name="sessions-per-second" value="30"/> Best regards / Mit freundlichen Grüßen, Daniel Neubert On 28.07.2010 08:54, Durmuş Ali Öztürk wrote:
I will try to set up a fs system on CentOS, maybe the performance will be better.
I am sorry but I dont understand the difference between cps and number of active session.
If fs has 10 active sessions (with media), then the call per second should be 10 too??
2010/7/28 Phillip Jones-2 [via freeswitch-users] <[hidden email] </user/SendEmail.jtp?type=node&node=5345283&i=0>>
>> The compiler flag is not found in my solution, should I get the latest source code
That would seem like a reasonable thing to do.....
>> In case of a timing problem, why it does not occured while calling parallel 30 number?
In my (limited) experience there are two important things to control:
1) call per second 2) Total number of active sessions
Calls per seconds appears (to me) to have a greater effect on the CPU. So at 1 call per second I would be looking at your CPU utilization and if it is maxing out then reducing that rate.
My point being here that your box *may* be capable of handling 100 or 1000 sessions, if the cps is lowered.
HTH
On Tue, Jul 27, 2010 at 2:39 AM, Durmuş Ali Öztürk <[hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=0>> wrote:
The compiler flag is not found in my solution, should I get the latest source code?
In case of a timing problem, why it does not occured while calling parallel 30 number? -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... <http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html?by-user=t&by-user=t> Sent from the freeswitch-users mailing list archive at Nabble.com.
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=1>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org <http://www.freeswitch.org?by-user=t&by-user=t>
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=2> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org <http://www.freeswitch.org?by-user=t&by-user=t&by-user=t>
------------------------------------------------------------------------ View message @ http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... <http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5344088.html?by-user=t>
To unsubscribe from Re: 100 parallel calls not possible (Performance problems), click here.
------------------------------------------------------------------------ View this message in context: Re: 100 parallel calls not possible (Performance problems) <http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5345283.html> Sent from the freeswitch-users mailing list archive <http://freeswitch-users.2379917.n2.nabble.com/> at Nabble.com.
Hallo Daniel, Yes, I have increased the value to 100. What are your limits and which system/platform do you use? MfG Ali 2010/7/28 Daniel Neubert <daniel.neubert@solomo.de>
Hi,
have you checked your switch.conf.xml configuration for "sessions-per-second" value?
<param name="sessions-per-second" value="30"/>
Best regards / Mit freundlichen Grüßen, Daniel Neubert
On 28.07.2010 08:54, Durmuş Ali Öztürk wrote:
I will try to set up a fs system on CentOS, maybe the performance will be better.
I am sorry but I dont understand the difference between cps and number of active session.
If fs has 10 active sessions (with media), then the call per second should be 10 too??
2010/7/28 Phillip Jones-2 [via freeswitch-users] <[hidden email]<http://user/SendEmail.jtp?type=node&node=5345283&i=0>
The compiler flag is not found in my solution, should I get the latest source code
That would seem like a reasonable thing to do.....
In case of a timing problem, why it does not occured while calling parallel 30 number?
In my (limited) experience there are two important things to control:
1) call per second 2) Total number of active sessions
Calls per seconds appears (to me) to have a greater effect on the CPU. So at 1 call per second I would be looking at your CPU utilization and if it is maxing out then reducing that rate.
My point being here that your box *may* be capable of handling 100 or 1000 sessions, if the cps is lowered.
HTH
On Tue, Jul 27, 2010 at 2:39 AM, Durmuş Ali Öztürk <[hidden email]<http://user/SendEmail.jtp?type=node&node=5344088&i=0>
wrote:
The compiler flag is not found in my solution, should I get the latest source code?
In case of a timing problem, why it does not occured while calling parallel 30 number? -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html<http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html?by-user=t&by-user=t> Sent from the freeswitch-users mailing list archive at Nabble.com.
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=1>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org<http://www.freeswitch.org?by-user=t&by-user=t>
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=2> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org<http://www.freeswitch.org?by-user=t&by-user=t&by-user=t>
------------------------------ View message @ http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5344088.html<http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5344088.html?by-user=t> To unsubscribe from Re: 100 parallel calls not possible (Performance problems), click here.
------------------------------ View this message in context: Re: 100 parallel calls not possible (Performance problems)<http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5345283.html>
Sent from the freeswitch-users mailing list archive<http://freeswitch-users.2379917.n2.nabble.com/>at Nabble.com.
We've not load-tested / stress-tested our platform setup, yet - but we will do this in a few weeks. Best regards / Mit freundlichen Grüßen, Daniel On 28.07.2010 11:35, Durmuş Ali Öztürk wrote:
Hallo Daniel,
Yes, I have increased the value to 100.
What are your limits and which system/platform do you use?
MfG Ali
2010/7/28 Daniel Neubert <daniel.neubert@solomo.de <mailto:daniel.neubert@solomo.de>>
Hi,
have you checked your switch.conf.xml configuration for "sessions-per-second" value?
<param name="sessions-per-second" value="30"/>
Best regards / Mit freundlichen Grüßen, Daniel Neubert
On 28.07.2010 08:54, Durmuş Ali Öztürk wrote:
I will try to set up a fs system on CentOS, maybe the performance will be better.
I am sorry but I dont understand the difference between cps and number of active session.
If fs has 10 active sessions (with media), then the call per second should be 10 too??
2010/7/28 Phillip Jones-2 [via freeswitch-users] <[hidden email] <http://user/SendEmail.jtp?type=node&node=5345283&i=0>>
>> The compiler flag is not found in my solution, should I get the latest source code
That would seem like a reasonable thing to do.....
>> In case of a timing problem, why it does not occured while calling parallel 30 number?
In my (limited) experience there are two important things to control:
1) call per second 2) Total number of active sessions
Calls per seconds appears (to me) to have a greater effect on the CPU. So at 1 call per second I would be looking at your CPU utilization and if it is maxing out then reducing that rate.
My point being here that your box *may* be capable of handling 100 or 1000 sessions, if the cps is lowered.
HTH
On Tue, Jul 27, 2010 at 2:39 AM, Durmuş Ali Öztürk <[hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=0>> wrote:
The compiler flag is not found in my solution, should I get the latest source code?
In case of a timing problem, why it does not occured while calling parallel 30 number? -- View this message in context: http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... <http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5340996.html?by-user=t&by-user=t> Sent from the freeswitch-users mailing list archive at Nabble.com.
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=1>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org <http://www.freeswitch.org?by-user=t&by-user=t>
_______________________________________________ FreeSWITCH-users mailing list [hidden email] <http://user/SendEmail.jtp?type=node&node=5344088&i=2> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org <http://www.freeswitch.org?by-user=t&by-user=t&by-user=t>
------------------------------------------------------------------------ View message @ http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possibl... <http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5344088.html?by-user=t>
To unsubscribe from Re: 100 parallel calls not possible (Performance problems), click here.
------------------------------------------------------------------------ View this message in context: Re: 100 parallel calls not possible (Performance problems) <http://freeswitch-users.2379917.n2.nabble.com/100-parallel-calls-not-possible-Performance-problems-tp5339029p5345283.html>
Sent from the freeswitch-users mailing list archive <http://freeswitch-users.2379917.n2.nabble.com/> at Nabble.com.
participants (6)
-
Daniel Neubert
-
Durmuş Ali Öztürk
-
Jay Binks
-
Jeff Lenk
-
Nyamul Hassan
-
Phillip Jones