For no apparent reason (under great network conditions), some attendees (but not all) in a multi-party meeting (e.g.,7 people) will see everyone in tile-mode freeze except for the dominant speaker. When someone new becomes the dominant speaker, the new dominant speaker starts moving and everyone else is frozen.
It is as if LastN was set to 1, but we have LastN hardcoded to 16 for all meeting sizes.
We aren’t seeing this if we use VP8, but we can’t use VP8 because VP8 doesn’t handle packet loss nearly as well as VP9.
It feels as though we’re missing some setting – unless we’re hitting a bug in VP9’s layer-forwarding or BWE-respect logic.
What would be the suggestion as to how to enable VP9 right now? I know it’s not officially supported yet, but there’s a “preferredcodec” flag in config that allows you to specify VP9 (and VP9 shows up as codec used in the meeting); then there’s the often-used method of setting it through sip communicator properties.
There was another thread that talked about how to force the bridge into using vp9. I believe my developers simply followed that thread. The only major bug we have found is the one I mentioned above, where sometimes only the active speaker has an active video feed (simulating lastN=1). And since we don’t know how to fix it, we don’t actually know if it’s related to vp9, BWE or something else.
My developers tell me we’re using the latest stable (5142). And they said the settings are as follows:
Add following to /etc/jitsi/jicofo/sip-communicator.properties
org.jitsi.jicofo.ENABLE_H264=false
org.jitsi.jicofo.ENABLE_VP8=false
org.jitsi.jicofo.ENABLE_VP9=true
Add following to /etc/jitsi/videobridge/sip-communicator.properties
org.jitsi.videobridge.ENABLE_SVC=true
org.jitsi.videobridge.ENABLE_VP9_SVC=true
Note: As I am not a developer, I cannot confirm this! I’m just passing it on.