We now have tested MCU 2.0 version in our service pruffme.com in webinars. And we noticed that quality of video (from publisher) changes during videoconference vie time. Is this changing of quality of video depend from Internet traffic and MCU regulates this quality?
We have tried testing the webinars with 2 speakers and 50 participants.
When only I speakers in webinar is speaking – it is all good! Sounds and video from speaker is good!
But when two speakers simultaneously are speaking - the sounds and video is too bad and there is a delay and it is not possible to hear speakers normally by participants.
That is a reason of this?
We noticed the next things:
1) when one publisher:
every 2-3 minutes decreases quality of video. It becomes really terrible. Voice is ok that time. Then video quality becomes good for some time and next all repeats.
2) when two publishers in one room(and about 30-40 viewers):
all begins well, but in 30-60 seconds both voice and video start works worth and worth. An then translating(or recieveing) stops. Only refresh of page helps.
Logs of translating time are in zip449795
All these happened in your scenarios are the QoS functionality we provide in MCU server, as mentioned in our 2.0 release: "Added the following intelligent QoS controls (NACK, FEC, and dynamic bitrate control), to improve conference quality against packet loss and network bandwidth variance.".
Generally, MCU will decrease the quality of video output when it detects the network bandwidth on server is not enough. And, most case of bad audio output quality is due to packet loss.
So first thing is that you need to make sure the network bandwidth of MCU server is large enough for your scenario. Otherwise you would experience quality vibration. If you can monitor the network status and throughput on MCU, especially for each publisher and subscriber, upload those data and we can do further analysis, so that we might be able to refine our QoS algorithm if all possible.
As I mentioned in my Email, really great to see it work on pruffme.com. Let's make it better! Thanks!
We have a server with unlimited brandwidth. With minimum of 200Mb till 1Gb.
In out tests with 130 participants the brandwidth was 13Mb. But described above problem happend even with 1mb brandwidth.
Also when we had 2 speakers - one of them came with good quality and another came with awful quality, thow the internet was ok.
As we seen - quality changed for output streams, and not depends on input streams.
And still the problem exists, as I wrote in anothe brunch:
Works the same way without any difference. It would be better to change it ourselfs, not by server.
Is there any way to disable dynamic bitrate control?
We have only bmon installed. What program can log information that will be enough for you to understand problems?
Currently the QoS policy on MCU cannot be disabled.
It seems your network condition is very good. We're trying to figure out a better solution and some possible algorithm refinement might be needed as well.
Also, we noticed some fact:
First - the quality can be good. But internrt connection of publisher(ore reciever) is not stable. And when speed becomes slow - the quality of video decreases. But whern internet speed becomes good again - the quality of video doesn't increase.
So even internet connection is good, but not stable all the time - we will have bad video.
I think mechanizm should increase quality of video ofter instenet speed becomes good again
The current design of QoS in forward mode heavily depends on the client connection to MCU which has the worst network connection. We will definitely refine this strategy in our future release, or at lease let yourself to decide whether to enable QoS on MCU or not.