[jitsi-dev] Videobridge performance test suite and forcing Jitsi Meet to use PCMA


#1

Hello,

Is there a test suite which allows us to test how the Videobridge scales? For example, checking how delays, throughput, packet loss and other metrics behave with increasing traffic. I saw the jitsi-meet-torture repository, but I don't think it has these kind of tests.

I was also wondering if it's possible to force clients to use PCMA in Jitsi Meet? I am measuring how much latency transcoding at Jigasi adds, and would like to compare the results when there is no transcoding (our SIP trunk supports PCMA only).

Regards,
Markus


#2

Hi Markus,

Hello,

Is there a test suite which allows us to test how the Videobridge scales? For example, checking how delays, throughput, packet loss and other metrics behave with increasing traffic. I saw the jitsi-meet-torture repository, but I don�t think it has these kind of tests.

There isn't a complete solution. We used jitsi-hammer in the past, but have transitioned to using selenium based clients because it is now easier to run many machines in the cloud and the traffic is more representative of real conferences.

I was also wondering if it�s possible to force clients to use PCMA in Jitsi Meet? I am measuring how much latency transcoding at Jigasi adds, and would like to compare the results when there is no transcoding (our SIP trunk supports PCMA only).

Even if clients send PCMA jigasi will have to transcode and mix before sending to the SIP trunk (right Damencho?). I'm not sure if advertising PCMA on the XMPP side will avoid transcoding, but you can do so by adding it here:
https://github.com/jitsi/jicofo/blob/master/src/main/java/org/jitsi/jicofo/util/JingleOfferFactory.java#L589

I hope that helps,
Boris

···

On 24/05/2018 00:04, markus.kullberg@teliacompany.com wrote:


#3

If Jicofo advertises PCMA, you can switch Jigasi to translator mode to
avoid transcoding and mixing.

/Kaiduan

···

On Thu, May 24, 2018 at 11:51 AM, Boris Grozev <boris@jitsi.org> wrote:

Hi Markus,

On 24/05/2018 00:04, markus.kullberg@teliacompany.com wrote:

Hello,

Is there a test suite which allows us to test how the Videobridge scales?
For example, checking how delays, throughput, packet loss and other metrics
behave with increasing traffic. I saw the jitsi-meet-torture repository,
but I don’t think it has these kind of tests.

There isn't a complete solution. We used jitsi-hammer in the past, but
have transitioned to using selenium based clients because it is now easier
to run many machines in the cloud and the traffic is more representative of
real conferences.

I was also wondering if it’s possible to force clients to use PCMA in
Jitsi Meet? I am measuring how much latency transcoding at Jigasi adds, and
would like to compare the results when there is no transcoding (our SIP
trunk supports PCMA only).

Even if clients send PCMA jigasi will have to transcode and mix before
sending to the SIP trunk (right Damencho?). I'm not sure if advertising
PCMA on the XMPP side will avoid transcoding, but you can do so by adding
it here:
https://github.com/jitsi/jicofo/blob/master/src/main/java/
org/jitsi/jicofo/util/JingleOfferFactory.java#L589

I hope that helps,
Boris

_______________________________________________
dev mailing list
dev@jitsi.org
Unsubscribe instructions and other list options:
http://lists.jitsi.org/mailman/listinfo/dev

--
Founder of Goodstartsoft
https://www.goodstartsoft.com


#4

In translator mode you also need support for it on the sip side, cause it
will start just forwarding all streams with different ssrc and will expect
the sip side to support and mix it.
So in not translator mode all streams are decoded and mixed in jigasi, no
matter of the codec.

···

On Thu, May 24, 2018, 11:31 Kaiduan Xie <kaiduanx@gmail.com> wrote:

If Jicofo advertises PCMA, you can switch Jigasi to translator mode to
avoid transcoding and mixing.

/Kaiduan

On Thu, May 24, 2018 at 11:51 AM, Boris Grozev <boris@jitsi.org> wrote:

Hi Markus,

On 24/05/2018 00:04, markus.kullberg@teliacompany.com wrote:

Hello,

Is there a test suite which allows us to test how the Videobridge
scales? For example, checking how delays, throughput, packet loss and other
metrics behave with increasing traffic. I saw the jitsi-meet-torture
repository, but I don’t think it has these kind of tests.

There isn't a complete solution. We used jitsi-hammer in the past, but
have transitioned to using selenium based clients because it is now easier
to run many machines in the cloud and the traffic is more representative of
real conferences.

I was also wondering if it’s possible to force clients to use PCMA in
Jitsi Meet? I am measuring how much latency transcoding at Jigasi adds, and
would like to compare the results when there is no transcoding (our SIP
trunk supports PCMA only).

Even if clients send PCMA jigasi will have to transcode and mix before
sending to the SIP trunk (right Damencho?). I'm not sure if advertising
PCMA on the XMPP side will avoid transcoding, but you can do so by adding
it here:

https://github.com/jitsi/jicofo/blob/master/src/main/java/org/jitsi/jicofo/util/JingleOfferFactory.java#L589

I hope that helps,
Boris

_______________________________________________
dev mailing list
dev@jitsi.org
Unsubscribe instructions and other list options:
http://lists.jitsi.org/mailman/listinfo/dev

--
Founder of Goodstartsoft
https://www.goodstartsoft.com
_______________________________________________
dev mailing list
dev@jitsi.org
Unsubscribe instructions and other list options:
http://lists.jitsi.org/mailman/listinfo/dev


#5

Hi,

Thanks for your replies. I will give the translator mode a try.

Boris: Would it be possible for you to share a bit more information about the testing process? I am only just learning Selenium and would appreciate any tips. For example, are you using Selenium Grid and/or Docker or any other tools worth mentioning? Do you create 1 virtual machine for every “test user” and distribute the VMs over different availability zones? How do you gather statistics from the individual conferences?’

Sorry for the many questions :slight_smile:

Regards,
Markus

···

From: dev [mailto:dev-bounces@jitsi.org] On Behalf Of Damian Minkov
Sent: 24. toukokuuta 2018 19:36
To: Jitsi Developers <dev@jitsi.org>
Subject: Re: [jitsi-dev] Videobridge performance test suite and forcing Jitsi Meet to use PCMA

In translator mode you also need support for it on the sip side, cause it will start just forwarding all streams with different ssrc and will expect the sip side to support and mix it.
So in not translator mode all streams are decoded and mixed in jigasi, no matter of the codec.

On Thu, May 24, 2018, 11:31 Kaiduan Xie <kaiduanx@gmail.com<mailto:kaiduanx@gmail.com>> wrote:
If Jicofo advertises PCMA, you can switch Jigasi to translator mode to avoid transcoding and mixing.

/Kaiduan

On Thu, May 24, 2018 at 11:51 AM, Boris Grozev <boris@jitsi.org<mailto:boris@jitsi.org>> wrote:
Hi Markus,

On 24/05/2018 00:04, markus.kullberg@teliacompany.com<mailto:markus.kullberg@teliacompany.com> wrote:
Hello,

Is there a test suite which allows us to test how the Videobridge scales? For example, checking how delays, throughput, packet loss and other metrics behave with increasing traffic. I saw the jitsi-meet-torture repository, but I don’t think it has these kind of tests.

There isn't a complete solution. We used jitsi-hammer in the past, but have transitioned to using selenium based clients because it is now easier to run many machines in the cloud and the traffic is more representative of real conferences.

I was also wondering if it’s possible to force clients to use PCMA in Jitsi Meet? I am measuring how much latency transcoding at Jigasi adds, and would like to compare the results when there is no transcoding (our SIP trunk supports PCMA only).

Even if clients send PCMA jigasi will have to transcode and mix before sending to the SIP trunk (right Damencho?). I'm not sure if advertising PCMA on the XMPP side will avoid transcoding, but you can do so by adding it here:
https://github.com/jitsi/jicofo/blob/master/src/main/java/org/jitsi/jicofo/util/JingleOfferFactory.java#L589

I hope that helps,
Boris

_______________________________________________
dev mailing list
dev@jitsi.org<mailto:dev@jitsi.org>
Unsubscribe instructions and other list options:
http://lists.jitsi.org/mailman/listinfo/dev

--
Founder of Goodstartsoft
https://www.goodstartsoft.com
_______________________________________________
dev mailing list
dev@jitsi.org<mailto:dev@jitsi.org>
Unsubscribe instructions and other list options:
http://lists.jitsi.org/mailman/listinfo/dev


#6

We are wrapping up some tests now and will publish details soon (probably next week).

Regards,
Boris

···

On 24/05/2018 12:47, markus.kullberg@teliacompany.com wrote:

Hi,

Thanks for your replies. I will give the translator mode a try.

Boris: Would it be possible for you to share a bit more information about the testing process? I am only just learning Selenium and would appreciate any tips. For example, are you using Selenium Grid and/or Docker or any other tools worth mentioning? Do you create 1 virtual machine for every “test user” and distribute the VMs over different availability zones? How do you gather statistics from the individual conferences?’