Multiple WebRTC questions.

Skip to first unread message

Alex Bondarenko

May 20, 2016, 11:29:59 AM5/20/16
to discuss-webrtc
We are developing cloud service for Live and VOD streaming for IP/WEB/Smartphone cameras. 
And we are trying to build our system completely on open-source solutions.
Currently we are using HLS, MPEG-DASH and Flash.
Recently we've stared research on WebRTC as real time capabilities are great. 
Usually we grab RTSP h.264 mux it to MP4 and play it in the browsers via HTML5.
Usually 1 person at a time is watching VOD and up to 10 persons watching Live.
We have to grab RTSP from the camera by our cloud server.
Then on demand, we are starting to stream Live or VOD to our customer from our cloud server. 
As i understand, if we grab h.264 and play h.264, there's no transcoding, but SVC. 
How many resources consume SVC in comparison to transcoding?
Is there any expamples of WebRTC working with multiple file(segments) video archives with seamless transition between files(segments) in the player?
What of these services support SVC, SFU, MCU and their combinations?
Intel WebRTC SDK
Can i somewhere get performance characteristics like this: for the services listed above?

Thanks in advance, would appreciate answers on any of the questions above. 


May 20, 2016, 3:40:46 PM5/20/16
there is no SVC support in webrtc.
webrtc only works with full, real-time, streams not segments of pre-encoded videos (like HLS).

you can use whatever you want on the back end or as a second "peer" (SFU or MCU, with normal or SVC codecs), as long as they support eh corresponding IETF specs for codecs, encryption, .....

intel WebRTC solution include an hybrid SFU/MCU, and a SIP gateway, on the server side
kurento is a mixer (I think)
Jitsi is an SFU, with support for simulcast, written in JAVA
licode is an SFU, written mainly in C++
meedoze I don t remember
Janus is an SFU by default, but can become much more depending on the "plugin" you choose. lightweight C.

You did not list MediaSoup, but you might want to take a look at it. 

There is no common benchmark for MCU/SFUs. We had a lot of discussion on the list in the past, and the most difficult is to decide your use case before you do any benchmark. As a result, many have provided numbers, but which are not really comparable as-is.

Possible use cases of interest:

- Max users in a conference
-- all at fixed resolution, and all rendered
-- simulcast (one rendered full res, the others at lower res)
-- last N (only the last active are rendered)

- maximum size of a conference in a unique server
-- fixed reference hardware
-- possible limit on bandwidth

- maximum number of supported user across a platform

and that s only for the conference case, hen there is broadcast, .... and so on an so forth.


You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to
To view this discussion on the web visit
For more options, visit

Alex. Gouaillard, PhD, PhD, MBA
Principal Architect - Citrix, San Francisco
President - CoSMo Software Consulting, Singapore

Luis Lopez

May 20, 2016, 4:20:03 PM5/20/16

El 20 may 2016, a las 21:40, Alexandre GOUAILLARD <> escribió:

kurento is a mixer (I think)

Alex, Kurento is not a mixer. Kurento is a modular media server and you, as a developer, may select the modules you wish to use. You have modules making Kurento to work as a mixer but you can also make it work as a media-switching-mixer (no decoding). It also has a SFU module with simulcast support.

By the way, Kurento is written in C/C++. It speaks a media server control protocol based on JSON, although there are Java and JavaScript SDKs for controlling it.




May 21, 2016, 11:03:05 PM5/21/16
thanks luis.


You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to

For more options, visit
Reply all
Reply to author
0 new messages