Does Video Rate Adaptation for video only lower resolution?

[Nadav]

Hi everyone,

I haven't been able to discern from the SRND and jabber guides how VRA dynamically allocates bandwidth. For example if I were to limit the bandwidth per endpoint (via QOS at the WAN) to be 1Mbps, would a RTCP enabled device start off with the lowest possible resolution and codec, and then ramp up until it finds the combination that is as close to 1Mbps as possible? Does it only affect resolution or also other factors? Can I see what was the final profile used when limited by the QoS? 

I'm using CUCM 11.0 and Jabber for Windows 11.0, as well as DX series endpoints.

Thanks! 

[Nadav]

Any ideas, guys?

[Mohammed al Baqari]

Hi,

RTCP will start from high to low under it adjusts to the available bandwidth. This will reflect on the resolution of the video stream. 

You can't find this from the jabber client.

[Nadav]

I see, so it only lowers the resolution and nothing else including voice codec? Is there any way to configure the starting resolution of the jabber/video client so that there aren't spikes in WAN bandwidth?

[Mohammed al Baqari]

You can limit the start bandwidth using video CAC in CUCM.
 

[Nadav]

From what I recall, CAC applies to the total bandwidth allocated in a location and not per endpoint. This means that the first 30 could start out at full bandwidth and the 31st person would start out with less since they have to share the bandwidth with everyone else in the location. It would be better to avoid bandwidth spikes by making each user begin with a lower resolution and not try the highest resolution and work down from there.

Is there a way to make everyone start out at 480p, for example?

[Mohammed al Baqari]

Hi,

Let me detail more. You can use RSVP based CAC. In RSVP config you can specific the max bandwidth per session to avoid spikes. You can use software MTP resources rather than hardware to act as RSVP agents. In this case, you video bandwidth will be capped and no spikes will be present.

[Nadav]

If all video-enabled SIP devices register to the same CUCM 11.x cluster, would it be possible to define max bandwidth per session between each location? All the examples I've seen for RSVP require all the L3 hardware between the endpoints to be configured as an RSVP agent. Have you seen documentation which shows how to configure RSVP based CAC so that the entire configuration is within the CUCM cluster?

Thanks for your time, I appreciate it :)

[Nadav]

I spoke with someone who integrates UC solutions regularly, and he believes that you can configure a maximum video bandwidth per call via the Region. Does this ring true?

[Mohammed al Baqari]

Not really. This can only be done using RSVP. I will be interested to hear from him how this can be done.

[Dennis Mink · Accepted solution]

That is correct. 

If you use the region settings for Video and you use SIP for your endpoints, you will notice that in the SIP SDP's it will use TIAS values in the signalling that coincide with whatever you configured in your region settings. so lets say your region A>B is defined to use 2Mbps for video (which is in fact audio+video+sometimes evene presentation), your endpoint will kick off with that 2Mbps in its signalling.  I all depends on the far end if that is eventually what will be transmitted.