Freeswitch memory issue/question
Dear developers: Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests. Here is the test senario: Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% . Please help Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code: switch_core_session_execute_application(session, "bridge", "user/500001"); Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan... your input would be very much appreciated. thanks -- Henry Huang aka bbhenry VoIP & Open Source software Consultant
We have a policy against load testing issues on this mailing list. What you are doing is most likely doing something wrong and we don't have time to debug it for you. We get way too many requests like this and there is a 99% result in proving user error and improper load testing. You can consider paid support from FreeSWITCH Solutions LLC for $175/hr per consultant and an ongoing support contract for $10k/yr to deal with business related bugfixing and support. The free help we give here is based on a community shared resource pool and we simply do not have enough to go around helping people with your type of issue. Please feel free to report bugs if you can pinpoint one using a more realistic use case that does not involve DDosing your box with 10 million invites. On Wed, Apr 14, 2010 at 5:11 AM, Henry Huang <b_ball_henry@hotmail.com>wrote:
Dear developers:
Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests.
Here is the test senario:
Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% .
Please help
Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code:
switch_core_session_execute_application(session, "bridge", "user/500001");
Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan...
your input would be very much appreciated.
thanks
-- Henry Huang aka bbhenry
VoIP & Open Source software Consultant
_______________________________________________ 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
-- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900
Henry Huang, with 10 millions of invites I'm sure you have at least $175 to pay to FS to help you !! :D ----- Original Message ----- From: Anthony Minessale To: freeswitch-users@lists.freeswitch.org Sent: Wednesday, April 14, 2010 11:19 AM Subject: Re: [Freeswitch-users] Freeswitch memory issue/question We have a policy against load testing issues on this mailing list. What you are doing is most likely doing something wrong and we don't have time to debug it for you. We get way too many requests like this and there is a 99% result in proving user error and improper load testing. You can consider paid support from FreeSWITCH Solutions LLC for $175/hr per consultant and an ongoing support contract for $10k/yr to deal with business related bugfixing and support. The free help we give here is based on a community shared resource pool and we simply do not have enough to go around helping people with your type of issue. Please feel free to report bugs if you can pinpoint one using a more realistic use case that does not involve DDosing your box with 10 million invites. On Wed, Apr 14, 2010 at 5:11 AM, Henry Huang <b_ball_henry@hotmail.com> wrote: Dear developers: Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests. Here is the test senario: Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% . Please help Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code: switch_core_session_execute_application(session, "bridge", "user/500001"); Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan... your input would be very much appreciated. thanks -- Henry Huang aka bbhenry VoIP & Open Source software Consultant _______________________________________________ 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 -- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire AIM: anthm MSN:anthony_minessale@hotmail.com GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org googletalk:conf+888@conference.freeswitch.org pstn:+19193869900 ------------------------------------------------------------------------------ _______________________________________________ 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
Henry's 10 millions of INVITEs are from a sipp test scenario, not from live commercial traffic. On 14/4/10 6:35 ??, Madovsky wrote:
Henry Huang, with 10 millions of invites I'm sure you have at least $175 to pay to FS to help you !! :D
----- Original Message ----- *From:* Anthony Minessale <mailto:anthony.minessale@gmail.com> *To:* freeswitch-users@lists.freeswitch.org <mailto:freeswitch-users@lists.freeswitch.org> *Sent:* Wednesday, April 14, 2010 11:19 AM *Subject:* Re: [Freeswitch-users] Freeswitch memory issue/question
We have a policy against load testing issues on this mailing list.
What you are doing is most likely doing something wrong and we don't have time to debug it for you. We get way too many requests like this and there is a 99% result in proving user error and improper load testing.
You can consider paid support from FreeSWITCH Solutions LLC for $175/hr per consultant and an ongoing support contract for $10k/yr to deal with business related bugfixing and support.
The free help we give here is based on a community shared resource pool and we simply do not have enough to go around helping people with your type of issue. Please feel free to report bugs if you can pinpoint one using a more realistic use case that does not involve DDosing your box with 10 million invites.
On Wed, Apr 14, 2010 at 5:11 AM, Henry Huang <b_ball_henry@hotmail.com <mailto:b_ball_henry@hotmail.com>> wrote:
Dear developers:
Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests.
Here is the test senario:
Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% .
Please help
Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code:
switch_core_session_execute_application(session, "bridge", "user/500001");
Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan...
your input would be very much appreciated.
thanks
-- Henry Huang aka bbhenry
VoIP & Open Source software Consultant
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org <mailto: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
-- Anthony Minessale II
FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire
AIM: anthm MSN:anthony_minessale@hotmail.com <mailto:MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com <mailto:PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net <http://irc.freenode.net> #freeswitch
FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <mailto:sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org <mailto:googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900
------------------------------------------------------------------------ _______________________________________________ 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
_______________________________________________ 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
ho yes ? I didn't know ;) ----- Original Message ----- From: Vlasis Hatzistavrou (KTI) To: freeswitch-users@lists.freeswitch.org Sent: Wednesday, April 14, 2010 1:20 PM Subject: Re: [Freeswitch-users] Freeswitch memory issue/question Henry's 10 millions of INVITEs are from a sipp test scenario, not from live commercial traffic. On 14/4/10 6:35 ??, Madovsky wrote: Henry Huang, with 10 millions of invites I'm sure you have at least $175 to pay to FS to help you !! :D ----- Original Message ----- From: Anthony Minessale To: freeswitch-users@lists.freeswitch.org Sent: Wednesday, April 14, 2010 11:19 AM Subject: Re: [Freeswitch-users] Freeswitch memory issue/question We have a policy against load testing issues on this mailing list. What you are doing is most likely doing something wrong and we don't have time to debug it for you. We get way too many requests like this and there is a 99% result in proving user error and improper load testing. You can consider paid support from FreeSWITCH Solutions LLC for $175/hr per consultant and an ongoing support contract for $10k/yr to deal with business related bugfixing and support. The free help we give here is based on a community shared resource pool and we simply do not have enough to go around helping people with your type of issue. Please feel free to report bugs if you can pinpoint one using a more realistic use case that does not involve DDosing your box with 10 million invites. On Wed, Apr 14, 2010 at 5:11 AM, Henry Huang <b_ball_henry@hotmail.com> wrote: Dear developers: Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests. Here is the test senario: Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% . Please help Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code: switch_core_session_execute_application(session, "bridge", "user/500001"); Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan... your input would be very much appreciated. thanks -- Henry Huang aka bbhenry VoIP & Open Source software Consultant _______________________________________________ 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 -- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire AIM: anthm MSN:anthony_minessale@hotmail.com GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org googletalk:conf+888@conference.freeswitch.org pstn:+19193869900 -------------------------------------------------------------------------- _______________________________________________ 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 _______________________________________________ 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 ------------------------------------------------------------------------------ _______________________________________________ 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
Henry, Realistically - what is your expected traffic? i.e. how big a community are you building this system for? In your test - now many cps where you hitting FS with? On Wed, Apr 14, 2010 at 11:35 AM, Madovsky <infos@madovsky.org> wrote:
Henry Huang,
with 10 millions of invites I'm sure you have at least $175 to pay to FS to help you !! :D
----- Original Message ----- *From:* Anthony Minessale <anthony.minessale@gmail.com> *To:* freeswitch-users@lists.freeswitch.org *Sent:* Wednesday, April 14, 2010 11:19 AM *Subject:* Re: [Freeswitch-users] Freeswitch memory issue/question
We have a policy against load testing issues on this mailing list.
What you are doing is most likely doing something wrong and we don't have time to debug it for you. We get way too many requests like this and there is a 99% result in proving user error and improper load testing.
You can consider paid support from FreeSWITCH Solutions LLC for $175/hr per consultant and an ongoing support contract for $10k/yr to deal with business related bugfixing and support.
The free help we give here is based on a community shared resource pool and we simply do not have enough to go around helping people with your type of issue. Please feel free to report bugs if you can pinpoint one using a more realistic use case that does not involve DDosing your box with 10 million invites.
On Wed, Apr 14, 2010 at 5:11 AM, Henry Huang <b_ball_henry@hotmail.com>wrote:
Dear developers:
Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests.
Here is the test senario:
Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% .
Please help
Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code:
switch_core_session_execute_application(session, "bridge", "user/500001");
Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan...
your input would be very much appreciated.
thanks
-- Henry Huang aka bbhenry
VoIP & Open Source software Consultant
_______________________________________________ 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
-- Anthony Minessale II
FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire
AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch
FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900
------------------------------
_______________________________________________ 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
_______________________________________________ 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
1 hint, you need enough ram to hold on to every dialog for several minutes even after the call has ended per the RFC so you will see a large increase in memory if you send millions of calls at the box. On Wed, Apr 14, 2010 at 12:26 PM, Phillip Jones <pjintheusa@gmail.com>wrote:
Henry,
Realistically - what is your expected traffic? i.e. how big a community are you building this system for? In your test - now many cps where you hitting FS with?
On Wed, Apr 14, 2010 at 11:35 AM, Madovsky <infos@madovsky.org> wrote:
Henry Huang,
with 10 millions of invites I'm sure you have at least $175 to pay to FS to help you !! :D
----- Original Message ----- *From:* Anthony Minessale <anthony.minessale@gmail.com> *To:* freeswitch-users@lists.freeswitch.org *Sent:* Wednesday, April 14, 2010 11:19 AM *Subject:* Re: [Freeswitch-users] Freeswitch memory issue/question
We have a policy against load testing issues on this mailing list.
What you are doing is most likely doing something wrong and we don't have time to debug it for you. We get way too many requests like this and there is a 99% result in proving user error and improper load testing.
You can consider paid support from FreeSWITCH Solutions LLC for $175/hr per consultant and an ongoing support contract for $10k/yr to deal with business related bugfixing and support.
The free help we give here is based on a community shared resource pool and we simply do not have enough to go around helping people with your type of issue. Please feel free to report bugs if you can pinpoint one using a more realistic use case that does not involve DDosing your box with 10 million invites.
On Wed, Apr 14, 2010 at 5:11 AM, Henry Huang <b_ball_henry@hotmail.com>wrote:
Dear developers:
Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests.
Here is the test senario:
Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% .
Please help
Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code:
switch_core_session_execute_application(session, "bridge", "user/500001");
Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan...
your input would be very much appreciated.
thanks
-- Henry Huang aka bbhenry
VoIP & Open Source software Consultant
_______________________________________________ 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
-- Anthony Minessale II
FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire
AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch
FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900
------------------------------
_______________________________________________ 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
_______________________________________________ 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
_______________________________________________ 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
-- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900
Anthony: yeh, I thought it could be something wrong with my testing setup too. And you are right, these kind of question requires professional support. Thanks for taking the time to reply me. But if anyone in the community would be kind to point out what should be a more proper way of conducting a SIPp test based on their experience, I am all ears. Thanks again, and FS 1.0.6 is truly great. Jumping from 1.0.4 to 1.0.6, we found it more robust and bug free. Henry On Wed, Apr 14, 2010 at 11:19 PM, Anthony Minessale < anthony.minessale@gmail.com> wrote:
We have a policy against load testing issues on this mailing list.
What you are doing is most likely doing something wrong and we don't have time to debug it for you. We get way too many requests like this and there is a 99% result in proving user error and improper load testing.
You can consider paid support from FreeSWITCH Solutions LLC for $175/hr per consultant and an ongoing support contract for $10k/yr to deal with business related bugfixing and support.
The free help we give here is based on a community shared resource pool and we simply do not have enough to go around helping people with your type of issue. Please feel free to report bugs if you can pinpoint one using a more realistic use case that does not involve DDosing your box with 10 million invites.
On Wed, Apr 14, 2010 at 5:11 AM, Henry Huang <b_ball_henry@hotmail.com>wrote:
Dear developers:
Our product is close to sipping stage and we are doing stress tests on FS servers. We have now hit a dead end with our test results and running out of ideas. And the result is pointing to memory issues. We have ran our test on both FS 1.0.4 and 1.0.6 trunk with SIPP to do simply invite tests.
Here is the test senario:
Send 10,000,000 invites to our test server running FS 1.0.6 with fixed amount of invites per second. The xml dialplan simply bridge the call to an unknown/unregistered user like so (user/500001) and therefore FS would hang up once it doesn't find the user. When running the test, FS is steady, but the memory usage slowly piles up over time. The memory keeps adding up till it hit the ceiling, then FS crash. I don't understand what could be holding the memory and not releasing it since we are not doing anything special and the calls are not being answered. And I forgot to mention, through out the whole test, the CPU usage stays at about 35% .
Please help
Besides FS native test, we have created our own application module to run the same test. The app is written in C and basically what we did is to point the xml dialplan to the custom app. Inside our app , we are simply doing the same kind of test as the xml dialplan would do. The only thing we did for the stress test is the following code:
switch_core_session_execute_application(session, "bridge", "user/500001");
Like the xml dialplan test, we are bridging the call to a unregistered user, so the call would hang up right away if system doesn't find the user. But the bizarre thing is this - the memory usage is about 7 ~ 8 times as much as we are running only xml dialplan. This I don't understand. Our app is calling the native FS C function to bridge the call , why would it use more memory than the xml dialplan...
your input would be very much appreciated.
thanks
-- Henry Huang aka bbhenry
VoIP & Open Source software Consultant
_______________________________________________ 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
-- Anthony Minessale II
FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire
AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch
FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900
_______________________________________________ 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
-- Henry Huang VoIP & Open Source software Consultant
participants (6)
-
Anthony Minessale
-
Henry Huang
-
Henry Huang
-
Madovsky
-
Phillip Jones
-
Vlasis Hatzistavrou (KTI)