Yes, we could use one encoder for all the participants, but then you
couldn't have different views per participant, or have different
encoding parameters per participant or perform bit rate adaptation per
participant (planed).
Currently the architecture has been thought to be able to be as flexible
as possible.. so it could be possible in the same conference to have an
H264 participant in HD at 2MBs watching 16 people at the same time, and
participants in a mobile phone using qcif and h263 at 128kbs watching
only the hd one.. for example.. but that comes with a cost. Anyway, I
expect anyone deploying that kind of service to be able to afford buying
a nice server or cluster or servers.. :)
Best regards
Sergio
El 30/05/2011 14:21, Kiran N escribió:
Hi Sergio,
Thanks for the quick response. I was able to reduce the CPU utilization to
6% (when only one participant is present in the conference) by performing the
following
1. Re-compile the x264 module with yasm support.
2. Changed to QCIF format
3. FPS is reduced to 10.
The reason for why I am sticking to H.264 is that, I am testing the application
with both Eyebeam and SIP
Communicator. For some reason, SIP Communicator is
not able to understand the H.263-1998 format from the MCU and thus nothing is
shown
on the
screen.
I may be wrong, I had a peek into the code and thus came to know that
a VideoEncoder object is made for each participant and thus when there are 2
participants
the CPU usage is 12% (6 + 6) and when its 3 participants the usage is 18%.
This might be a stupid question, but it came to my mind
Isn't it possible for us to have a single VideoEncoder per conference?
Thanks and Regards,
Kiran
>Hi Kiran,
>The CPU utilization is something that we cannot reduce, it is bound to
>the encoding process, which is something expensive if using h264.
>You can do one of the following things, reduce the video quality (either
>going to QCIF, or H263 or reducing fps) or increase the CPU power of
>your server. I have made test with up to 8 participants using HD(720P)
>video at 30fps.
>Also, the mcu allows you have more than one mixer configured to be able
>to use more than one server at a time.
>Best regards
>Sergio
--- On*Mon, 30/5/11, Kiran N/<[email protected]>/* wrote:
From: Kiran N <[email protected]>
Subject: CPU Utilization of H.264 Conference with MCU
To: [email protected]
Date: Monday, 30 May, 2011, 12:56 PM
Hi -,
I am trying to use 2X2 conference with H.264 Codec with CIF
format. With a single person in the conference i could see that
the CPU utilization has gone to 25% and when the second fellow
joins, it went up to 50%. I could see that H.264 encoder (x264
lib) is the one responsible for most of the CPU utilization.
Does this mean that I would not be able to run more that 2-3
conferences (say 2X2) active on a system? Is there any way, we can
reduce the CPU usage?
Any help in this regard will be greatly appreciated.
Thanks and Regards,
Kiran
--
_____________________________________________________________________
-- Bandwidth and Colocation Provided by http://www.api-digital.com --
asterisk-video mailing list
To UNSUBSCRIBE or update options visit:
http://lists.digium.com/mailman/listinfo/asterisk-video
--
_____________________________________________________________________
-- Bandwidth and Colocation Provided by http://www.api-digital.com --
asterisk-video mailing list
To UNSUBSCRIBE or update options visit:
http://lists.digium.com/mailman/listinfo/asterisk-video
