Hardware transcoding solutions?
We have several existing solutions for 729<->711 and DTMF transcoding, but they are either horribly expensive and not very scalable from a licensing perspective, or are chewing up a buttload of power in our facilities. What is the current "best option" out there for a 1 to 4 U hardware (ASIC driven?) transcoding solution? I've looked at the Sangoma NetBorder box already. Brooks Bridges | Sr. Voice Services Engineer O1 Communications 5190 Golden Foothill Pkwy El Dorado Hills, CA 95762 office: 916.235.2097 | main: 888.444.1111, Option 2 email: bbridges at o1.com<mailto:bbridges at o1.com> | web: www.o1.com<http://www.o1.com/>
Nothing at all? No one here does hardware transcoding outside their existing SBCs? Brooks Bridges | Sr. Voice Services Engineer O1 Communications 5190 Golden Foothill Pkwy El Dorado Hills, CA 95762 office: 916.235.2097 | main: 888.444.1111, Option 2 email: bbridges at o1.com<mailto:bbridges at o1.com> | web: www.o1.com<http://www.o1.com/> From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Brooks Bridges Sent: Monday, August 17, 2015 3:47 PM To: voiceops at voiceops.org Subject: [VoiceOps] Hardware transcoding solutions? We have several existing solutions for 729<->711 and DTMF transcoding, but they are either horribly expensive and not very scalable from a licensing perspective, or are chewing up a buttload of power in our facilities. What is the current "best option" out there for a 1 to 4 U hardware (ASIC driven?) transcoding solution? I've looked at the Sangoma NetBorder box already. Brooks Bridges | Sr. Voice Services Engineer O1 Communications 5190 Golden Foothill Pkwy El Dorado Hills, CA 95762 office: 916.235.2097 | main: 888.444.1111, Option 2 email: bbridges at o1.com<mailto:bbridges at o1.com> | web: www.o1.com<http://www.o1.com/>
Le 2015-08-27 15:10, Brooks Bridges a ?crit :
We have several existing solutions for 729<->711 and DTMF transcoding, but they are either horribly expensive and not very scalable from a licensing perspective, or are chewing up a buttload of power in our facilities.
What is the current ?best option? out there for a 1 to 4 U hardware (ASIC driven?) transcoding solution? I?ve looked at the Sangoma NetBorder box already.
How many simultaneous calls? Simon
Our total requirement is > 5000 channels, however I'm more than willing to deploy multiple devices to reach this number as long as the power and space requirements aren't horribly restrictive. Brooks Bridges | Sr. Voice Services Engineer O1 Communications 5190 Golden Foothill Pkwy El Dorado Hills, CA 95762 office: 916.235.2097 | main: 888.444.1111, Option 2 email: bbridges at o1.com |?web: www.o1.com -----Original Message----- From: Simon Perreault [mailto:sperreault at jive.com] Sent: Thursday, August 27, 2015 12:28 PM To: Brooks Bridges; voiceops at voiceops.org Subject: Re: [VoiceOps] Hardware transcoding solutions? Le 2015-08-27 15:10, Brooks Bridges a ?crit :
We have several existing solutions for 729<->711 and DTMF transcoding, but they are either horribly expensive and not very scalable from a licensing perspective, or are chewing up a buttload of power in our facilities.
What is the current "best option" out there for a 1 to 4 U hardware (ASIC driven?) transcoding solution? I've looked at the Sangoma NetBorder box already.
How many simultaneous calls? Simon
Le 2015-08-27 15:38, Brooks Bridges a ?crit :
Our total requirement is > 5000 channels, however I'm more than willing to deploy multiple devices to reach this number as long as the power and space requirements aren't horribly restrictive.
One can easily transcode 5000 729<->711 channels in software on a generic 1U server. Have you even tried? Simon
On Thu, Aug 27, 2015 at 3:45 PM, Simon Perreault <sperreault at jive.com> wrote:
Le 2015-08-27 15:38, Brooks Bridges a ?crit :
Our total requirement is > 5000 channels, however I'm more than willing to deploy multiple devices to reach this number as long as the power and space requirements aren't horribly restrictive.
One can easily transcode 5000 729<->711 channels in software on a generic 1U server. Have you even tried?
I'm curious. Can you be more precise on what you mean exactly by generic 1U server? What matters mostly is the processor, not the rack unit, so I'd be interested in hearing your experience with transcoding 5000 g729 to g711 channels. Here are some precise numbers that I run on a dual xeon 1U server earlier this month for a presentation at the FreeSWITCH conference, ClueCon. The tests were run using 1,000 calls at 50cps (using FreeSWITCH). However, I did test at lower cps, same max calls and results are quite similar because the bulk of the load is the transcoding as it can be seen in the graphs below. Without transcoding (just PCMU to PCMU bridging with standard 20ms packetization): http://moythreads.com/vmstat/xeon-sqlitefullmem-lwarn-sprof-glibc-1000 at 50cps/report.html And with G.722 to PCMU transcoding: http://www.moythreads.com/vmstat/xeon-sqlitefullmem-lwarn-sprof-glibc-transc... at 50cps/report.html At 1,000 calls you already maxed the CPUs with G.722 transcoding, see the last graph, that's what's more relevant, showing how the idle cpu on the server went from 80% to 15% idle on average just when adding transcoding. Granted, there's more powerful boxes out there, but this should give people a clear idea of the cost of transcoding. The algorithmic complexity of G722 is on par with G.729, in fact, G.729 is a bit more complex. Not to mention other codecs such as iLBC or AMR. Overall, I agree software can give very decent results, so it's really a matter of using what you see fits better your needs. - Moy
Le 2015-08-27 17:00, Moises Silva a ?crit :
I'm curious. Can you be more precise on what you mean exactly by generic 1U server? What matters mostly is the processor, not the rack unit, so I'd be interested in hearing your experience with transcoding 5000 g729 to g711 channels.
I'm sorry, I can't, for obvious reasons.
Here are some precise numbers that I run on a dual xeon 1U server earlier this month for a presentation at the FreeSWITCH conference, ClueCon.
The tests were run using 1,000 calls at 50cps (using FreeSWITCH). However, I did test at lower cps, same max calls and results are quite similar because the bulk of the load is the transcoding
That is my experience as well.
as it can be seen in the graphs below.
Without transcoding (just PCMU to PCMU bridging with standard 20ms packetization):
http://moythreads.com/vmstat/xeon-sqlitefullmem-lwarn-sprof-glibc-1000 at 50cps/report.html
And with G.722 to PCMU transcoding:
http://www.moythreads.com/vmstat/xeon-sqlitefullmem-lwarn-sprof-glibc-transc... at 50cps/report.html
At 1,000 calls you already maxed the CPUs with G.722 transcoding, see the last graph, that's what's more relevant, showing how the idle cpu on the server went from 80% to 15% idle on average just when adding transcoding. Granted, there's more powerful boxes out there, but this should give people a clear idea of the cost of transcoding.
...on FreeSWITCH. I'd be curious to see the output of "perf top", to see if CPU time is being spent wisely.
The algorithmic complexity of G722 is on par with G.729, in fact, G.729 is a bit more complex.
That is my experience as well. Simon
On Thu, Aug 27, 2015 at 5:09 PM, Simon Perreault <sperreault at jive.com> wrote:
Le 2015-08-27 17:00, Moises Silva a ?crit :
I'm curious. Can you be more precise on what you mean exactly by generic 1U server? What matters mostly is the processor, not the rack unit, so I'd be interested in hearing your experience with transcoding 5000 g729 to g711 channels.
I'm sorry, I can't, for obvious reasons.
I guess that does not help the OP, since he can't get a hold of your specialized software nor have any hints on the hardware specs you used to achieve that :)
Le 2015-08-27 17:13, Moises Silva a ?crit :
On Thu, Aug 27, 2015 at 5:09 PM, Simon Perreault <sperreault at jive.com <mailto:sperreault at jive.com>> wrote:
Le 2015-08-27 17:00, Moises Silva a ?crit : > I'm curious. Can you be more precise on what you mean exactly by generic > 1U server? What matters mostly is the processor, not the rack unit, so > I'd be interested in hearing your experience with transcoding 5000 g729 > to g711 channels.
I'm sorry, I can't, for obvious reasons.
I guess that does not help the OP, since he can't get a hold of your specialized software nor have any hints on the hardware specs you used to achieve that :)
I know, and I'm truly sorry for that. I wish I could go into more details, it really is fascinating (at least to me!). The hardware is generic and I don't see any reason FreeSWITCH/Asterisk/anything else couldn't achieve the same performance. Hence my question: "Have you even tried?" I would have liked the OP to post some benchmarking results to start the discussion, but it looks as though he just assumed that software couldn't possibly attain his required performance level. If that is true, then providing a rough idea of what level of performance software can attain *is* indeed helping the OP. Simon
I have no interest in building anything in-house for this. I'm only interested in a properly supported and reliable commercial grade solution. Brooks Bridges | Sr. Voice Services Engineer O1 Communications 5190 Golden Foothill Pkwy El Dorado Hills, CA 95762 office: 916.235.2097 | main: 888.444.1111, Option 2 email: bbridges at o1.com |?web: www.o1.com -----Original Message----- From: Simon Perreault [mailto:sperreault at jive.com] Sent: Thursday, August 27, 2015 2:22 PM To: Moises Silva Cc: Brooks Bridges; voiceops at voiceops.org Subject: Re: [VoiceOps] Hardware transcoding solutions? Le 2015-08-27 17:13, Moises Silva a ?crit :
On Thu, Aug 27, 2015 at 5:09 PM, Simon Perreault <sperreault at jive.com <mailto:sperreault at jive.com>> wrote:
Le 2015-08-27 17:00, Moises Silva a ?crit : > I'm curious. Can you be more precise on what you mean exactly by generic > 1U server? What matters mostly is the processor, not the rack unit, so > I'd be interested in hearing your experience with transcoding 5000 g729 > to g711 channels.
I'm sorry, I can't, for obvious reasons.
I guess that does not help the OP, since he can't get a hold of your specialized software nor have any hints on the hardware specs you used to achieve that :)
I know, and I'm truly sorry for that. I wish I could go into more details, it really is fascinating (at least to me!). The hardware is generic and I don't see any reason FreeSWITCH/Asterisk/anything else couldn't achieve the same performance. Hence my question: "Have you even tried?" I would have liked the OP to post some benchmarking results to start the discussion, but it looks as though he just assumed that software couldn't possibly attain his required performance level. If that is true, then providing a rough idea of what level of performance software can attain *is* indeed helping the OP. Simon
Le 2015-08-27 17:23, Brooks Bridges a ?crit :
I have no interest in building anything in-house for this. I'm only interested in a properly supported and reliable commercial grade solution.
Right, and I'm not suggesting that. What I've come to understand is that by "hardware solution" you meant "a box you can buy" as opposed to a home-grown solution, not as opposed to a software solution. And that makes sense to me. Simon
On 08/27/2015 05:21 PM, Simon Perreault wrote:
The hardware is generic and I don't see any reason FreeSWITCH/Asterisk/anything else couldn't achieve the same performance. Hence my question: "Have you even tried?" I would have liked the OP to post some benchmarking results to start the discussion, but it looks as though he just assumed that software couldn't possibly attain his required performance level. If that is true, then providing a rough idea of what level of performance software can attain *is* indeed helping the OP.
Based on Brooks' post, I think the physical and power footprint requirements as well as in-house supportability are paramount here. There's no question that a few Freeswitch boxes can do this in software just fine, and I don't think that's under controversy. But I don't think it's what he wants. Simon, I think you're under the impression that you're arguing with a traditional ILEC blockhead/troglodyte who doesn't understand FOSS. That's an understandable assumption given the amount of people on this list who seem to be constitutionally unable/who cholerically refuse to grasp that one doesn't need to spend $200,000 on a proprietary appliance to accomplish fairly trivial objectives, but in this case incorrect. It's just not what Brooks wants. :-) -- Alex Balashov | Principal | Evariste Systems LLC 303 Perimeter Center North, Suite 300 Atlanta, GA 30346 United States Tel: +1-800-250-5920 (toll-free) / +1-678-954-0671 (direct) Web: http://www.evaristesys.com/, http://www.csrpswitch.com/
Le 2015-08-27 17:31, Alex Balashov a ?crit :
Simon, I think you're under the impression that you're arguing
I was under the impression we were having a nice friendly chat! :)
with a traditional ILEC blockhead/troglodyte who doesn't understand FOSS. That's an understandable assumption given the amount of people on this list who seem to be constitutionally unable/who cholerically refuse to grasp that one doesn't need to spend $200,000 on a proprietary appliance to accomplish fairly trivial objectives, but in this case incorrect. It's just not what Brooks wants. :-)
I have never argued the business aspect of all this. I just pointed out that software *can* do 5k calls in one 1U generic server. I'm not saying the OP *should* do it, I just don't know enough about his situation. Simon
On 08/27/2015 05:35 PM, Simon Perreault wrote:
I just pointed out that software*can* do 5k calls in one 1U generic server
5K of transcoded calls, with a growth path? Admittedly, it's based on anecdata rather than an enquiry of any scientific rigour, but my experiences with Freeswitch point to results closer to those of Moises Silva. It seems that it can do 1000-2000 transcoded channels on a commodity multi-core box without a problem. However, it's been a few years, so it's possible that the density is greater today, and/or that Freeswitch can fully and efficiently utilise more and faster cores nowadays to achieve 4000-5000. -- Alex Balashov | Principal | Evariste Systems LLC 303 Perimeter Center North, Suite 300 Atlanta, GA 30346 United States Tel: +1-800-250-5920 (toll-free) / +1-678-954-0671 (direct) Web: http://www.evaristesys.com/, http://www.csrpswitch.com/
Le 2015-08-27 17:38, Alex Balashov a ?crit :
On 08/27/2015 05:35 PM, Simon Perreault wrote:
I just pointed out that software*can* do 5k calls in one 1U generic server
5K of transcoded calls, with a growth path?
Sure, the growth path is Moore's law. Just buy more and bigger servers. That's the beauty of software.
Admittedly, it's based on anecdata rather than an enquiry of any scientific rigour, but my experiences with Freeswitch point to results closer to those of Moises Silva. It seems that it can do 1000-2000 transcoded channels on a commodity multi-core box without a problem. However, it's been a few years, so it's possible that the density is greater today, and/or that Freeswitch can fully and efficiently utilise more and faster cores nowadays to achieve 4000-5000.
I honestly have no idea of FreeSWITCH's performance so can't talk about it. :) Simon
I would say whatever solution you pick, make sure it is able to transcode OPUS. That would be a requirement for me. On Thu, Aug 27, 2015 at 4:41 PM, Simon Perreault <sperreault at jive.com> wrote:
Le 2015-08-27 17:38, Alex Balashov a ?crit :
On 08/27/2015 05:35 PM, Simon Perreault wrote:
I just pointed out that software*can* do 5k calls in one 1U generic server
5K of transcoded calls, with a growth path?
Sure, the growth path is Moore's law. Just buy more and bigger servers. That's the beauty of software.
Admittedly, it's based on anecdata rather than an enquiry of any scientific rigour, but my experiences with Freeswitch point to results closer to those of Moises Silva. It seems that it can do 1000-2000 transcoded channels on a commodity multi-core box without a problem. However, it's been a few years, so it's possible that the density is greater today, and/or that Freeswitch can fully and efficiently utilise more and faster cores nowadays to achieve 4000-5000.
I honestly have no idea of FreeSWITCH's performance so can't talk about it.
:)
Simon _______________________________________________ VoiceOps mailing list VoiceOps at voiceops.org https://puck.nether.net/mailman/listinfo/voiceops
On Thu, Aug 27, 2015 at 5:35 PM, Simon Perreault <sperreault at jive.com> wrote:
Le 2015-08-27 17:31, Alex Balashov a ?crit :
Simon, I think you're under the impression that you're arguing
I was under the impression we were having a nice friendly chat! :)
Nonsense, there's no such thing on mailing lists :)
Cisco AS5400 can do this right? On Aug 27, 2015 15:10, Brooks Bridges <bbridges at o1.com> wrote:
Nothing at all?? No one here does hardware transcoding outside their existing SBCs?
?
Brooks Bridges | Sr. Voice Services Engineer
O1 Communications
5190 Golden Foothill Pkwy
El Dorado Hills, CA 95762
office: 916.235.2097 | main: 888.444.1111, Option 2
email: bbridges at o1.com |?web: www.o1.com
?
From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Brooks Bridges Sent: Monday, August 17, 2015 3:47 PM To: voiceops at voiceops.org Subject: [VoiceOps] Hardware transcoding solutions?
?
We have several existing solutions for 729<->711 and DTMF transcoding, but they are either horribly expensive and not very scalable from a licensing perspective, or are chewing up a buttload of power in our facilities.
?
What is the current ?best option? out there for a 1 to 4 U hardware (ASIC driven?) transcoding solution?? I?ve looked at the Sangoma NetBorder box already.
?
Brooks Bridges | Sr. Voice Services Engineer
O1 Communications
5190 Golden Foothill Pkwy
El Dorado Hills, CA 95762
office: 916.235.2097 | main: 888.444.1111, Option 2
email: bbridges at o1.com |?web: www.o1.com
?
It does, but I don't think it would get us the density we're shooting for. Brooks Bridges | Sr. Voice Services Engineer O1 Communications 5190 Golden Foothill Pkwy El Dorado Hills, CA 95762 office: 916.235.2097 | main: 888.444.1111, Option 2 email: bbridges at o1.com |?web: www.o1.com -----Original Message----- From: Paul Timmins [mailto:paul at timmins.net] Sent: Thursday, August 27, 2015 12:38 PM To: Brooks Bridges Cc: voiceops at voiceops.org Subject: Re: [VoiceOps] Hardware transcoding solutions? Cisco AS5400 can do this right? On Aug 27, 2015 15:10, Brooks Bridges <bbridges at o1.com> wrote:
Nothing at all?? No one here does hardware transcoding outside their existing SBCs?
?
Brooks Bridges | Sr. Voice Services Engineer
O1 Communications
5190 Golden Foothill Pkwy
El Dorado Hills, CA 95762
office: 916.235.2097 | main: 888.444.1111, Option 2
email: bbridges at o1.com |?web: www.o1.com
?
From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Brooks Bridges Sent: Monday, August 17, 2015 3:47 PM To: voiceops at voiceops.org Subject: [VoiceOps] Hardware transcoding solutions?
?
We have several existing solutions for 729<->711 and DTMF transcoding, but they are either horribly expensive and not very scalable from a licensing perspective, or are chewing up a buttload of power in our facilities.
?
What is the current ?best option? out there for a 1 to 4 U hardware (ASIC driven?) transcoding solution?? I?ve looked at the Sangoma NetBorder box already.
?
Brooks Bridges | Sr. Voice Services Engineer
O1 Communications
5190 Golden Foothill Pkwy
El Dorado Hills, CA 95762
office: 916.235.2097 | main: 888.444.1111, Option 2
email: bbridges at o1.com |?web: www.o1.com
?
Check Sangoma, they had a variety of sizes and flavors last time I checked. Did not get a chance to use it however, so can't say anything good or bad. On 15-08-27 03:10 PM, Brooks Bridges wrote:
Nothing at all? No one here does hardware transcoding outside their existing SBCs?
*Brooks Bridges | *Sr. Voice Services Engineer
*O^1 Communications*
5190 Golden Foothill Pkwy
El Dorado Hills, CA 95762
*office:*916.235.2097 |*main:*888.444.1111, Option 2
*email:*bbridges at o1.com <mailto:bbridges at o1.com> | *web:*www.o1.com <http://www.o1.com/>
*From:*VoiceOps [mailto:voiceops-bounces at voiceops.org] *On Behalf Of *Brooks Bridges *Sent:* Monday, August 17, 2015 3:47 PM *To:* voiceops at voiceops.org *Subject:* [VoiceOps] Hardware transcoding solutions?
We have several existing solutions for 729<->711 and DTMF transcoding, but they are either horribly expensive and not very scalable from a licensing perspective, or are chewing up a buttload of power in our facilities.
What is the current ?best option? out there for a 1 to 4 U hardware (ASIC driven?) transcoding solution? I?ve looked at the Sangoma NetBorder box already.
*Brooks Bridges | *Sr. Voice Services Engineer
*O^1 Communications*
5190 Golden Foothill Pkwy
El Dorado Hills, CA 95762
*office:*916.235.2097 |*main:*888.444.1111, Option 2
*email:*bbridges at o1.com <mailto:bbridges at o1.com> | *web:*www.o1.com <http://www.o1.com/>
_______________________________________________ VoiceOps mailing list VoiceOps at voiceops.org https://puck.nether.net/mailman/listinfo/voiceops
participants (7)
-
abalashov＠evaristesys.com -
bbridges＠o1.com -
colton.conor＠gmail.com -
moises.silva＠gmail.com -
paul＠timmins.net -
sperreault＠jive.com -
victor.chukalovskiy＠gmail.com