RE: [Asterisk-Users] Performance testing of asterisk
Hi Tom- I wrote that (rudimentary) Perl script last year to simulate traffic from one system to another, although it can also be used between spans on the same machine. It's much better to have the load generated on a separate system , for obvious reasons. A couple of things: the traffic generating spans should be set up as pri_net in zapata.conf. Also, you need an E1 crossover cable from sender to receiver. This is wired as: 1 -- 4 2 -- 5 4 -- 1 5 -- 2 I didn't see the rest of the thread here, but if your configuration involves transcoding, you need to build this in to your test too. I understand that there are software VoIP load generator programs that can do just this. If you are using T1's or E1's, I hope that my script will be useful in some way, as I found that there is a call setup/teardown load too when lots of calls are handled. Regards Scott Stingel Scott M. Stingel President, Emerging Voice Technology, Inc. Palo Alto California & London England www.evtmedia.com -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Patrick Sent: Tuesday, August 17, 2004 1:38 AM To: [EMAIL PROTECTED] Subject: Re: [Asterisk-Users] Performance testing of asterisk On Tue, 2004-08-17 at 02:17, Tom Masterson wrote: > What we are finding is that things work quite well with a small number > of users/agents and callers i.e 10 or less. However when we put the > stuff in production (normally changes to the configurations) where > they can and are hit with hundreds of callers among some 60 or 70 > agents we have major failures. What we are trying to do is come up > with some automated way of creating the same effect so we can test > changes before they go out to the rest of the group. The generator > does not have to be the same box as the test asterisk box but we need > to recreate the scenarios as much as possible. > Scott [forgot surname] from Emerging Technologies (search the list messages from 2003) sent the attached script to the list during a discussion about stress testing asterisk. Iirc it works by looping 2 T1/E1/PRI ports on the same box and sending calls back and forth. Think it was possible also to loop calls between 2 asterisk boxes. Hope it is helpful. Regards, Patrick ___ Asterisk-Users mailing list [EMAIL PROTECTED] http://lists.digium.com/mailman/listinfo/asterisk-users To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [Asterisk-Users] Performance testing of asterisk
On Tue, 2004-08-17 at 02:17, Tom Masterson wrote: > What we are finding is that things work quite well with a small number of > users/agents and callers i.e 10 or less. However when we put the stuff in > production (normally changes to the configurations) where they can and are > hit with hundreds of callers among some 60 or 70 agents we have major > failures. What we are trying to do is come up with some automated way of > creating the same effect so we can test changes before they go out to the > rest of the group. The generator does not have to be the same box as the > test asterisk box but we need to recreate the scenarios as much as > possible. > Scott [forgot surname] from Emerging Technologies (search the list messages from 2003) sent the attached script to the list during a discussion about stress testing asterisk. Iirc it works by looping 2 T1/E1/PRI ports on the same box and sending calls back and forth. Think it was possible also to loop calls between 2 asterisk boxes. Hope it is helpful. Regards, Patrick call_generator.tar.gz Description: application/compressed-tar
Re: [Asterisk-Users] Performance testing of asterisk
What we are finding is that things work quite well with a small number of users/agents and callers i.e 10 or less. However when we put the stuff in production (normally changes to the configurations) where they can and are hit with hundreds of callers among some 60 or 70 agents we have major failures. What we are trying to do is come up with some automated way of creating the same effect so we can test changes before they go out to the rest of the group. The generator does not have to be the same box as the test asterisk box but we need to recreate the scenarios as much as possible. Any ideas? I agree with Chris but it might still be better than nothing. Tom On Mon, 16 Aug 2004, Chris A. Icide wrote: > The problem with this kind of testing is that you are testing asterisk, > using asterisk as the test generator. > > When you start getting performance based artifacts, are you going to be > able to tell if it's from the sections of asterisk handling the call > transactions and media streams or from the sections that are generating the > calls. How much of the system resources are going to be used in the > generation of the calls, how does this affect asterisk. > > Is the program flow of internally generated calls exactly the same as the > program flow for the type of calls you are using as the scenario for the test? > > I think you will find that using the same asterisk box as testee and tester > will result in not being able to identify what is causing the > artifacts... In other words, your testing tool will significantly affect > your test platform. > > -Chris > > On 03:07 PM 8/16/2004, Tom Masterson wrote: > >We are trying to set up some scripts to test asterisk under various loads. > >What we are doing is trying to load a bunch of calls in to various queues > >atuomatically from various numbers etc so we can see how it behaves. I > >think we can do this by loading files in to the var/spool/asterisk/qcall > >directory. However the format of this file has a field named identifier > >which appears to be a file of some sort in some location. > > > >1. Are we going about this in the correct direction? > >2. Can someone explain what the fields in the file in qcall are and what > >identifier should be? > > > >Thanks > >Tom > > > >___ > >Asterisk-Users mailing list > >[EMAIL PROTECTED] > >http://lists.digium.com/mailman/listinfo/asterisk-users > >To UNSUBSCRIBE or update options visit: > > http://lists.digium.com/mailman/listinfo/asterisk-users > > ___ > Asterisk-Users mailing list > [EMAIL PROTECTED] > http://lists.digium.com/mailman/listinfo/asterisk-users > To UNSUBSCRIBE or update options visit: >http://lists.digium.com/mailman/listinfo/asterisk-users > ___ Asterisk-Users mailing list [EMAIL PROTECTED] http://lists.digium.com/mailman/listinfo/asterisk-users To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [Asterisk-Users] Performance testing of asterisk
The problem with this kind of testing is that you are testing asterisk, using asterisk as the test generator. When you start getting performance based artifacts, are you going to be able to tell if it's from the sections of asterisk handling the call transactions and media streams or from the sections that are generating the calls. How much of the system resources are going to be used in the generation of the calls, how does this affect asterisk. Is the program flow of internally generated calls exactly the same as the program flow for the type of calls you are using as the scenario for the test? I think you will find that using the same asterisk box as testee and tester will result in not being able to identify what is causing the artifacts... In other words, your testing tool will significantly affect your test platform. -Chris On 03:07 PM 8/16/2004, Tom Masterson wrote: >We are trying to set up some scripts to test asterisk under various loads. >What we are doing is trying to load a bunch of calls in to various queues >atuomatically from various numbers etc so we can see how it behaves. I >think we can do this by loading files in to the var/spool/asterisk/qcall >directory. However the format of this file has a field named identifier >which appears to be a file of some sort in some location. > >1. Are we going about this in the correct direction? >2. Can someone explain what the fields in the file in qcall are and what >identifier should be? > >Thanks >Tom > >___ >Asterisk-Users mailing list >[EMAIL PROTECTED] >http://lists.digium.com/mailman/listinfo/asterisk-users >To UNSUBSCRIBE or update options visit: > http://lists.digium.com/mailman/listinfo/asterisk-users ___ Asterisk-Users mailing list [EMAIL PROTECTED] http://lists.digium.com/mailman/listinfo/asterisk-users To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users