Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
Absolutely using wireshark to capture all RTP traffic for analysing is very challengeable because the size of capture file is very huge. So i have used -C option to spilite to multiple pcap file. I shared a picture of it: https://www.linkedin.com/posts/mojtaba-esfandiari_kamailio-kamailioworld-activity-6605747231586947072-M_x9 The result of you get is strange! On Tue, Jan 7, 2020 at 1:24 PM davy van de moere < davy.van.de.mo...@gmail.com> wrote: > Using wireshark, I get a jitter still smaller than 1ms. You have efficient > ways to measure that? > > Op di 7 jan. 2020 om 10:43 schreef Mojtaba : > >> Yes, I reached 3000 channels in kernel mode with about 25% usage of one >> core. Its incredible. But be notic the number of channels is not matter, >> What about its QoS parameter such as max-delta, jitter, lost? >> The max delay i have with the same result is 60. >> >> On Tue, Jan 7, 2020 at 12:56 PM davy van de moere < >> davy.van.de.mo...@gmail.com> wrote: >> >>> As an update, I'm getting 2000 channels in kernel mode, with about 50% >>> usage of 1 core. (load of the server is at 0.5), which is a 10 fold >>> difference. >>> >>> The problem I was hitting, was the (probable) inability to work in >>> kernelmode with the rtpproxy.so module, rtpengine.so works nicely. >>> >>> Op di 7 jan. 2020 om 10:19 schreef Mojtaba : >>> >>>> As a person who working much on developing new RTP engine for Kamailio, >>>> I should express achieving to 12k concurrent call is very strange with >>>> Kernel part of RTPEngine. >>>> Regardless of NIC's parameter, multi core and some other parameters, it >>>> is important where the hooking linux kernel function is called. The Kernel >>>> part of RTPEngine is called in NF_INET_LOCAL_IN. It means the rtp packet >>>> goes up local routing in network stack. >>>> I could not reach to 12k yet, although i call Linux kernel function in >>>> NF-INET-PRE-ROUTING >>>> >>>> On Mon, Jan 6, 2020 at 8:26 PM davy van de moere < >>>> davy.van.de.mo...@gmail.com> wrote: >>>> >>>>> It will probably be a case of RTFM... >>>>> >>>>> quoting from: https://github.com/sipwise/rtpengine/blob/master/README.md >>>>> >>>>> # ensure that the table we want to use doesn't exist - usually needed >>>>> after a daemon >>>>> # restart, otherwise will error >>>>> echo 'del 0' > /proc/rtpengine/control >>>>> >>>>> >>>>> this part seems to be very relevant indeed! >>>>> >>>>> Op ma 6 jan. 2020 om 17:41 schreef Nicolas Breuer < >>>>> nicolas.bre...@belcenter.biz>: >>>>> >>>>>> Hey Davy, >>>>>> >>>>>> >>>>>> >>>>>> Check cat /proc/rtpengine/0/list >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> *De :* sr-users *De la part de* >>>>>> davy van de moere >>>>>> *Envoyé :* lundi 6 janvier 2020 14:47 >>>>>> *À :* Henning Westerholt >>>>>> *Cc :* Kamailio (SER) - Users Mailing List < >>>>>> sr-users@lists.kamailio.org> >>>>>> *Objet :* Re: [SR-Users] RTPEngine performance - How far are you >>>>>> pushing your systems >>>>>> >>>>>> >>>>>> >>>>>> Ok, I'll get my SIPP warmed up then ... ;) thx for the feedback! >>>>>> >>>>>> >>>>>> >>>>>> Op ma 6 jan. 2020 om 14:45 schreef Henning Westerholt >>>>> >: >>>>>> >>>>>> Hello Davy, >>>>>> >>>>>> >>>>>> >>>>>> You should look into it, sounds strange. Compare e.g. to this e-mail >>>>>> from 2015 – over 16k simultaneous calls on a 12 core machine: >>>>>> >>>>>> https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html >>>>>> >>>>>> >>>>>> >>>>>> Cheers, >>>>>> >>>>>> >>>>>> >>>>>> Henning >>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>
Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
Have you checked /proc/rtpengine ? On Tue, 7 Jan 2020 at 09:56, davy van de moere wrote: > Using wireshark, I get a jitter still smaller than 1ms. You have efficient > ways to measure that? > > Op di 7 jan. 2020 om 10:43 schreef Mojtaba : > >> Yes, I reached 3000 channels in kernel mode with about 25% usage of one >> core. Its incredible. But be notic the number of channels is not matter, >> What about its QoS parameter such as max-delta, jitter, lost? >> The max delay i have with the same result is 60. >> >> On Tue, Jan 7, 2020 at 12:56 PM davy van de moere < >> davy.van.de.mo...@gmail.com> wrote: >> >>> As an update, I'm getting 2000 channels in kernel mode, with about 50% >>> usage of 1 core. (load of the server is at 0.5), which is a 10 fold >>> difference. >>> >>> The problem I was hitting, was the (probable) inability to work in >>> kernelmode with the rtpproxy.so module, rtpengine.so works nicely. >>> >>> Op di 7 jan. 2020 om 10:19 schreef Mojtaba : >>> >>>> As a person who working much on developing new RTP engine for Kamailio, >>>> I should express achieving to 12k concurrent call is very strange with >>>> Kernel part of RTPEngine. >>>> Regardless of NIC's parameter, multi core and some other parameters, it >>>> is important where the hooking linux kernel function is called. The Kernel >>>> part of RTPEngine is called in NF_INET_LOCAL_IN. It means the rtp packet >>>> goes up local routing in network stack. >>>> I could not reach to 12k yet, although i call Linux kernel function in >>>> NF-INET-PRE-ROUTING >>>> >>>> On Mon, Jan 6, 2020 at 8:26 PM davy van de moere < >>>> davy.van.de.mo...@gmail.com> wrote: >>>> >>>>> It will probably be a case of RTFM... >>>>> >>>>> quoting from: https://github.com/sipwise/rtpengine/blob/master/README.md >>>>> >>>>> # ensure that the table we want to use doesn't exist - usually needed >>>>> after a daemon >>>>> # restart, otherwise will error >>>>> echo 'del 0' > /proc/rtpengine/control >>>>> >>>>> >>>>> this part seems to be very relevant indeed! >>>>> >>>>> Op ma 6 jan. 2020 om 17:41 schreef Nicolas Breuer < >>>>> nicolas.bre...@belcenter.biz>: >>>>> >>>>>> Hey Davy, >>>>>> >>>>>> >>>>>> >>>>>> Check cat /proc/rtpengine/0/list >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> *De :* sr-users *De la part de* >>>>>> davy van de moere >>>>>> *Envoyé :* lundi 6 janvier 2020 14:47 >>>>>> *À :* Henning Westerholt >>>>>> *Cc :* Kamailio (SER) - Users Mailing List < >>>>>> sr-users@lists.kamailio.org> >>>>>> *Objet :* Re: [SR-Users] RTPEngine performance - How far are you >>>>>> pushing your systems >>>>>> >>>>>> >>>>>> >>>>>> Ok, I'll get my SIPP warmed up then ... ;) thx for the feedback! >>>>>> >>>>>> >>>>>> >>>>>> Op ma 6 jan. 2020 om 14:45 schreef Henning Westerholt >>>>> >: >>>>>> >>>>>> Hello Davy, >>>>>> >>>>>> >>>>>> >>>>>> You should look into it, sounds strange. Compare e.g. to this e-mail >>>>>> from 2015 – over 16k simultaneous calls on a 12 core machine: >>>>>> >>>>>> https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html >>>>>> >>>>>> >>>>>> >>>>>> Cheers, >>>>>> >>>>>> >>>>>> >>>>>> Henning >>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> >>>>>> Henning Westerholt – https://skalatan.de/blog/ >>>>>> >>>>>> Kamailio services – https://gilawa.com >>>>>> >>>>>> >>>>>> >>>>>> *From:* sr-users *On Behalf Of >>>>>> *davy van de moere >>>>>> *Se
Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
Using wireshark, I get a jitter still smaller than 1ms. You have efficient ways to measure that? Op di 7 jan. 2020 om 10:43 schreef Mojtaba : > Yes, I reached 3000 channels in kernel mode with about 25% usage of one > core. Its incredible. But be notic the number of channels is not matter, > What about its QoS parameter such as max-delta, jitter, lost? > The max delay i have with the same result is 60. > > On Tue, Jan 7, 2020 at 12:56 PM davy van de moere < > davy.van.de.mo...@gmail.com> wrote: > >> As an update, I'm getting 2000 channels in kernel mode, with about 50% >> usage of 1 core. (load of the server is at 0.5), which is a 10 fold >> difference. >> >> The problem I was hitting, was the (probable) inability to work in >> kernelmode with the rtpproxy.so module, rtpengine.so works nicely. >> >> Op di 7 jan. 2020 om 10:19 schreef Mojtaba : >> >>> As a person who working much on developing new RTP engine for Kamailio, >>> I should express achieving to 12k concurrent call is very strange with >>> Kernel part of RTPEngine. >>> Regardless of NIC's parameter, multi core and some other parameters, it >>> is important where the hooking linux kernel function is called. The Kernel >>> part of RTPEngine is called in NF_INET_LOCAL_IN. It means the rtp packet >>> goes up local routing in network stack. >>> I could not reach to 12k yet, although i call Linux kernel function in >>> NF-INET-PRE-ROUTING >>> >>> On Mon, Jan 6, 2020 at 8:26 PM davy van de moere < >>> davy.van.de.mo...@gmail.com> wrote: >>> >>>> It will probably be a case of RTFM... >>>> >>>> quoting from: https://github.com/sipwise/rtpengine/blob/master/README.md >>>> >>>> # ensure that the table we want to use doesn't exist - usually needed >>>> after a daemon >>>> # restart, otherwise will error >>>> echo 'del 0' > /proc/rtpengine/control >>>> >>>> >>>> this part seems to be very relevant indeed! >>>> >>>> Op ma 6 jan. 2020 om 17:41 schreef Nicolas Breuer < >>>> nicolas.bre...@belcenter.biz>: >>>> >>>>> Hey Davy, >>>>> >>>>> >>>>> >>>>> Check cat /proc/rtpengine/0/list >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> *De :* sr-users *De la part de* >>>>> davy van de moere >>>>> *Envoyé :* lundi 6 janvier 2020 14:47 >>>>> *À :* Henning Westerholt >>>>> *Cc :* Kamailio (SER) - Users Mailing List < >>>>> sr-users@lists.kamailio.org> >>>>> *Objet :* Re: [SR-Users] RTPEngine performance - How far are you >>>>> pushing your systems >>>>> >>>>> >>>>> >>>>> Ok, I'll get my SIPP warmed up then ... ;) thx for the feedback! >>>>> >>>>> >>>>> >>>>> Op ma 6 jan. 2020 om 14:45 schreef Henning Westerholt >>>> >: >>>>> >>>>> Hello Davy, >>>>> >>>>> >>>>> >>>>> You should look into it, sounds strange. Compare e.g. to this e-mail >>>>> from 2015 – over 16k simultaneous calls on a 12 core machine: >>>>> >>>>> https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html >>>>> >>>>> >>>>> >>>>> Cheers, >>>>> >>>>> >>>>> >>>>> Henning >>>>> >>>>> >>>>> >>>>> -- >>>>> >>>>> Henning Westerholt – https://skalatan.de/blog/ >>>>> >>>>> Kamailio services – https://gilawa.com >>>>> >>>>> >>>>> >>>>> *From:* sr-users *On Behalf Of *davy >>>>> van de moere >>>>> *Sent:* Monday, January 6, 2020 12:52 PM >>>>> *To:* Kamailio (SER) - Users Mailing List >>>> > >>>>> *Subject:* [SR-Users] RTPEngine performance - How far are you pushing >>>>> your systems >>>>> >>>>> >>>>> >>>>> Dear all, >>>>> >>>>> >>>>> >>>>> Not a direct question on kamailio, but on rtpengine, so feel free to >>>>> ignor
Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
Yes, I reached 3000 channels in kernel mode with about 25% usage of one core. Its incredible. But be notic the number of channels is not matter, What about its QoS parameter such as max-delta, jitter, lost? The max delay i have with the same result is 60. On Tue, Jan 7, 2020 at 12:56 PM davy van de moere < davy.van.de.mo...@gmail.com> wrote: > As an update, I'm getting 2000 channels in kernel mode, with about 50% > usage of 1 core. (load of the server is at 0.5), which is a 10 fold > difference. > > The problem I was hitting, was the (probable) inability to work in > kernelmode with the rtpproxy.so module, rtpengine.so works nicely. > > Op di 7 jan. 2020 om 10:19 schreef Mojtaba : > >> As a person who working much on developing new RTP engine for Kamailio, I >> should express achieving to 12k concurrent call is very strange with Kernel >> part of RTPEngine. >> Regardless of NIC's parameter, multi core and some other parameters, it >> is important where the hooking linux kernel function is called. The Kernel >> part of RTPEngine is called in NF_INET_LOCAL_IN. It means the rtp packet >> goes up local routing in network stack. >> I could not reach to 12k yet, although i call Linux kernel function in >> NF-INET-PRE-ROUTING >> >> On Mon, Jan 6, 2020 at 8:26 PM davy van de moere < >> davy.van.de.mo...@gmail.com> wrote: >> >>> It will probably be a case of RTFM... >>> >>> quoting from: https://github.com/sipwise/rtpengine/blob/master/README.md >>> >>> # ensure that the table we want to use doesn't exist - usually needed after >>> a daemon >>> # restart, otherwise will error >>> echo 'del 0' > /proc/rtpengine/control >>> >>> >>> this part seems to be very relevant indeed! >>> >>> Op ma 6 jan. 2020 om 17:41 schreef Nicolas Breuer < >>> nicolas.bre...@belcenter.biz>: >>> >>>> Hey Davy, >>>> >>>> >>>> >>>> Check cat /proc/rtpengine/0/list >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>>> *De :* sr-users *De la part de* >>>> davy van de moere >>>> *Envoyé :* lundi 6 janvier 2020 14:47 >>>> *À :* Henning Westerholt >>>> *Cc :* Kamailio (SER) - Users Mailing List >>> > >>>> *Objet :* Re: [SR-Users] RTPEngine performance - How far are you >>>> pushing your systems >>>> >>>> >>>> >>>> Ok, I'll get my SIPP warmed up then ... ;) thx for the feedback! >>>> >>>> >>>> >>>> Op ma 6 jan. 2020 om 14:45 schreef Henning Westerholt : >>>> >>>> Hello Davy, >>>> >>>> >>>> >>>> You should look into it, sounds strange. Compare e.g. to this e-mail >>>> from 2015 – over 16k simultaneous calls on a 12 core machine: >>>> >>>> https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html >>>> >>>> >>>> >>>> Cheers, >>>> >>>> >>>> >>>> Henning >>>> >>>> >>>> >>>> -- >>>> >>>> Henning Westerholt – https://skalatan.de/blog/ >>>> >>>> Kamailio services – https://gilawa.com >>>> >>>> >>>> >>>> *From:* sr-users *On Behalf Of *davy >>>> van de moere >>>> *Sent:* Monday, January 6, 2020 12:52 PM >>>> *To:* Kamailio (SER) - Users Mailing List >>>> *Subject:* [SR-Users] RTPEngine performance - How far are you pushing >>>> your systems >>>> >>>> >>>> >>>> Dear all, >>>> >>>> >>>> >>>> Not a direct question on kamailio, but on rtpengine, so feel free to >>>> ignore :) >>>> >>>> >>>> >>>> I have a farm of rtpengine servers , supporting our set of kamailio's. >>>> >>>> >>>> >>>> On recent decent hardware (hp dl360 gen9/10 servers), I typically push >>>> my systems upto 2500 channels per machine. Then I tend to see the load of >>>> the machine go up in a logarithmic manner. E.g. a load of 3 supports 2500 >>>> channels, but when its at 3000 channels, load can be at 8. etc. >>>> >>>> >>>> >>>> I have the kernel module running, and I have a good impression it's >>>
Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
As an update, I'm getting 2000 channels in kernel mode, with about 50% usage of 1 core. (load of the server is at 0.5), which is a 10 fold difference. The problem I was hitting, was the (probable) inability to work in kernelmode with the rtpproxy.so module, rtpengine.so works nicely. Op di 7 jan. 2020 om 10:19 schreef Mojtaba : > As a person who working much on developing new RTP engine for Kamailio, I > should express achieving to 12k concurrent call is very strange with Kernel > part of RTPEngine. > Regardless of NIC's parameter, multi core and some other parameters, it is > important where the hooking linux kernel function is called. The Kernel > part of RTPEngine is called in NF_INET_LOCAL_IN. It means the rtp packet > goes up local routing in network stack. > I could not reach to 12k yet, although i call Linux kernel function in > NF-INET-PRE-ROUTING > > On Mon, Jan 6, 2020 at 8:26 PM davy van de moere < > davy.van.de.mo...@gmail.com> wrote: > >> It will probably be a case of RTFM... >> >> quoting from: https://github.com/sipwise/rtpengine/blob/master/README.md >> >> # ensure that the table we want to use doesn't exist - usually needed after >> a daemon >> # restart, otherwise will error >> echo 'del 0' > /proc/rtpengine/control >> >> >> this part seems to be very relevant indeed! >> >> Op ma 6 jan. 2020 om 17:41 schreef Nicolas Breuer < >> nicolas.bre...@belcenter.biz>: >> >>> Hey Davy, >>> >>> >>> >>> Check cat /proc/rtpengine/0/list >>> >>> >>> >>> >>> >>> >>> >>> *De :* sr-users *De la part de* >>> davy van de moere >>> *Envoyé :* lundi 6 janvier 2020 14:47 >>> *À :* Henning Westerholt >>> *Cc :* Kamailio (SER) - Users Mailing List >>> *Objet :* Re: [SR-Users] RTPEngine performance - How far are you >>> pushing your systems >>> >>> >>> >>> Ok, I'll get my SIPP warmed up then ... ;) thx for the feedback! >>> >>> >>> >>> Op ma 6 jan. 2020 om 14:45 schreef Henning Westerholt : >>> >>> Hello Davy, >>> >>> >>> >>> You should look into it, sounds strange. Compare e.g. to this e-mail >>> from 2015 – over 16k simultaneous calls on a 12 core machine: >>> >>> https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html >>> >>> >>> >>> Cheers, >>> >>> >>> >>> Henning >>> >>> >>> >>> -- >>> >>> Henning Westerholt – https://skalatan.de/blog/ >>> >>> Kamailio services – https://gilawa.com >>> >>> >>> >>> *From:* sr-users *On Behalf Of *davy >>> van de moere >>> *Sent:* Monday, January 6, 2020 12:52 PM >>> *To:* Kamailio (SER) - Users Mailing List >>> *Subject:* [SR-Users] RTPEngine performance - How far are you pushing >>> your systems >>> >>> >>> >>> Dear all, >>> >>> >>> >>> Not a direct question on kamailio, but on rtpengine, so feel free to >>> ignore :) >>> >>> >>> >>> I have a farm of rtpengine servers , supporting our set of kamailio's. >>> >>> >>> >>> On recent decent hardware (hp dl360 gen9/10 servers), I typically push >>> my systems upto 2500 channels per machine. Then I tend to see the load of >>> the machine go up in a logarithmic manner. E.g. a load of 3 supports 2500 >>> channels, but when its at 3000 channels, load can be at 8. etc. >>> >>> >>> >>> I have the kernel module running, and I have a good impression it's >>> being used too. No transcoding is going on, no recording, ... I've tried on >>> different OS'es, different network cards, and I always end up at that 2500 >>> channels +-. >>> >>> >>> >>> Is that what I can expect from rtpengine? Or am I really overlooking >>> something? What numbers do you push out of your systems? >>> >>> >>> >>> kind regards and happy 2020! >>> >>> ___ >>> Kamailio (SER) - Users Mailing List >>> sr-users@lists.kamailio.org >>> https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users >>> >> ___ >> Kamailio (SER) - Users Mailing List >> sr-users@lists.kamailio.org >> https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users >> > > > -- > --Mojtaba Esfandiari.S > ___ > Kamailio (SER) - Users Mailing List > sr-users@lists.kamailio.org > https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users > ___ Kamailio (SER) - Users Mailing List sr-users@lists.kamailio.org https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
It will probably be a case of RTFM... quoting from: https://github.com/sipwise/rtpengine/blob/master/README.md # ensure that the table we want to use doesn't exist - usually needed after a daemon # restart, otherwise will error echo 'del 0' > /proc/rtpengine/control this part seems to be very relevant indeed! Op ma 6 jan. 2020 om 17:41 schreef Nicolas Breuer < nicolas.bre...@belcenter.biz>: > Hey Davy, > > > > Check cat /proc/rtpengine/0/list > > > > > > > > *De :* sr-users *De la part de* > davy van de moere > *Envoyé :* lundi 6 janvier 2020 14:47 > *À :* Henning Westerholt > *Cc :* Kamailio (SER) - Users Mailing List > *Objet :* Re: [SR-Users] RTPEngine performance - How far are you pushing > your systems > > > > Ok, I'll get my SIPP warmed up then ... ;) thx for the feedback! > > > > Op ma 6 jan. 2020 om 14:45 schreef Henning Westerholt : > > Hello Davy, > > > > You should look into it, sounds strange. Compare e.g. to this e-mail from > 2015 – over 16k simultaneous calls on a 12 core machine: > > https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html > > > > Cheers, > > > > Henning > > > > -- > > Henning Westerholt – https://skalatan.de/blog/ > > Kamailio services – https://gilawa.com > > > > *From:* sr-users *On Behalf Of *davy > van de moere > *Sent:* Monday, January 6, 2020 12:52 PM > *To:* Kamailio (SER) - Users Mailing List > *Subject:* [SR-Users] RTPEngine performance - How far are you pushing > your systems > > > > Dear all, > > > > Not a direct question on kamailio, but on rtpengine, so feel free to > ignore :) > > > > I have a farm of rtpengine servers , supporting our set of kamailio's. > > > > On recent decent hardware (hp dl360 gen9/10 servers), I typically push my > systems upto 2500 channels per machine. Then I tend to see the load of the > machine go up in a logarithmic manner. E.g. a load of 3 supports 2500 > channels, but when its at 3000 channels, load can be at 8. etc. > > > > I have the kernel module running, and I have a good impression it's being > used too. No transcoding is going on, no recording, ... I've tried on > different OS'es, different network cards, and I always end up at that 2500 > channels +-. > > > > Is that what I can expect from rtpengine? Or am I really overlooking > something? What numbers do you push out of your systems? > > > > kind regards and happy 2020! > > ___ > Kamailio (SER) - Users Mailing List > sr-users@lists.kamailio.org > https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users > ___ Kamailio (SER) - Users Mailing List sr-users@lists.kamailio.org https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
Ok, I'll get my SIPP warmed up then ... ;) thx for the feedback! Op ma 6 jan. 2020 om 14:45 schreef Henning Westerholt : > Hello Davy, > > > > You should look into it, sounds strange. Compare e.g. to this e-mail from > 2015 – over 16k simultaneous calls on a 12 core machine: > > https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html > > > > Cheers, > > > > Henning > > > > -- > > Henning Westerholt – https://skalatan.de/blog/ > > Kamailio services – https://gilawa.com > > > > *From:* sr-users *On Behalf Of *davy > van de moere > *Sent:* Monday, January 6, 2020 12:52 PM > *To:* Kamailio (SER) - Users Mailing List > *Subject:* [SR-Users] RTPEngine performance - How far are you pushing > your systems > > > > Dear all, > > > > Not a direct question on kamailio, but on rtpengine, so feel free to > ignore :) > > > > I have a farm of rtpengine servers , supporting our set of kamailio's. > > > > On recent decent hardware (hp dl360 gen9/10 servers), I typically push my > systems upto 2500 channels per machine. Then I tend to see the load of the > machine go up in a logarithmic manner. E.g. a load of 3 supports 2500 > channels, but when its at 3000 channels, load can be at 8. etc. > > > > I have the kernel module running, and I have a good impression it's being > used too. No transcoding is going on, no recording, ... I've tried on > different OS'es, different network cards, and I always end up at that 2500 > channels +-. > > > > Is that what I can expect from rtpengine? Or am I really overlooking > something? What numbers do you push out of your systems? > > > > kind regards and happy 2020! > ___ Kamailio (SER) - Users Mailing List sr-users@lists.kamailio.org https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
Re: [SR-Users] RTPEngine performance - How far are you pushing your systems
Hello Davy, You should look into it, sounds strange. Compare e.g. to this e-mail from 2015 – over 16k simultaneous calls on a 12 core machine: https://lists.kamailio.org/pipermail/sr-dev/2015-October/031432.html Cheers, Henning -- Henning Westerholt – https://skalatan.de/blog/ Kamailio services – https://gilawa.com<https://gilawa.com/> From: sr-users On Behalf Of davy van de moere Sent: Monday, January 6, 2020 12:52 PM To: Kamailio (SER) - Users Mailing List Subject: [SR-Users] RTPEngine performance - How far are you pushing your systems Dear all, Not a direct question on kamailio, but on rtpengine, so feel free to ignore :) I have a farm of rtpengine servers , supporting our set of kamailio's. On recent decent hardware (hp dl360 gen9/10 servers), I typically push my systems upto 2500 channels per machine. Then I tend to see the load of the machine go up in a logarithmic manner. E.g. a load of 3 supports 2500 channels, but when its at 3000 channels, load can be at 8. etc. I have the kernel module running, and I have a good impression it's being used too. No transcoding is going on, no recording, ... I've tried on different OS'es, different network cards, and I always end up at that 2500 channels +-. Is that what I can expect from rtpengine? Or am I really overlooking something? What numbers do you push out of your systems? kind regards and happy 2020! ___ Kamailio (SER) - Users Mailing List sr-users@lists.kamailio.org https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
[SR-Users] RTPEngine performance - How far are you pushing your systems
Dear all, Not a direct question on kamailio, but on rtpengine, so feel free to ignore :) I have a farm of rtpengine servers , supporting our set of kamailio's. On recent decent hardware (hp dl360 gen9/10 servers), I typically push my systems upto 2500 channels per machine. Then I tend to see the load of the machine go up in a logarithmic manner. E.g. a load of 3 supports 2500 channels, but when its at 3000 channels, load can be at 8. etc. I have the kernel module running, and I have a good impression it's being used too. No transcoding is going on, no recording, ... I've tried on different OS'es, different network cards, and I always end up at that 2500 channels +-. Is that what I can expect from rtpengine? Or am I really overlooking something? What numbers do you push out of your systems? kind regards and happy 2020! ___ Kamailio (SER) - Users Mailing List sr-users@lists.kamailio.org https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
