within Pexip Infinity
The Pexip Infinity distributed architecture provides major bandwidth savings when compared with traditional MCU deployments.
When Conferencing Nodes that are hosting a conference (in the default 1 + 7 layout) communicate with each other over a backplane:
- The Conferencing Node to which the current speaker is connected sends an HD video stream of the current speaker, plus a lower-resolution thumbnail (a smaller image at the bottom of the screen which shows the participant's video) of each of the other participants connected to it.
- The Conferencing Node to which the previous speaker is connected sends an HD video stream of the previous speaker, plus lower-resolution thumbnails of each of the other participants connected to it.
- All other Conferencing Nodes send lower-resolution thumbnails of each of the participants connected to it.
As shown in the diagram above, if the current and previous speakers are connected to different Conferencing Nodes, there is one HD stream in each direction in the backplane between the two nodes, plus one smaller stream for every participant being shown in a thumbnail.
As shown in the diagram above, if the current and previous speakers are connected to the same Conferencing Node, that Conferencing Node sends an HD video stream of the current speaker plus lower-resolution thumbnails of each of the other participants, to all other Conferencing Nodes. The other Conferencing Nodes send lower-resolution thumbnails of the other participants.
This architecture means that there are major bandwidth savings when compared with traditional MCU deployments where all conference participants regardless of location connect to the same MCU and individual HD and thumbnail video streams are sent between the MCU and every endpoint.
The examples above are from the perspective of a conference using a "1 + 7" layout (1 large main speaker and up to 7 other thumbnail participants).
When using other layouts:
- In a 4 + 0 layout, when there are 1 or 2 participants to display, their streams are sent at high resolution over the backplane between nodes; if there are 3 or more participants then up to 5 streams are sent at a medium resolution.
- In a 2 + 21 layout, up to 3 HD streams may be sent between any 2 nodes (up to 2 current speakers and potentially 1 previous speaker).
- When using Adaptive Composition, different resolutions are sent as required for each position in the layout. Up to 13 streams could be sent between nodes: 3 top-row streams (one extra so that current speakers do not see themselves), 3 center-row streams and 7 bottom-row streams, all at their respective call qualities as required for the conference.
You should also note that:
- The bandwidth required by a HD stream on a backplane depends on a combination of many factors. Typically it uses about 1.6 Mbps but could use up to 4 Mbps.
- One thumbnail stream uses about 64-192 kbps.
- A presentation stream uses an additional 1.6 Mbps from the presenter.
- Pexip Infinity supports up to 6 Mbps per participant (this varies depending on the deployment environment, video resolutions, etc).
Pexip Infinity will automatically downspeed and upspeed individual calls in response to fluctuating network conditions.
Administrators can individually restrict the bandwidth available to participants accessing each Virtual Reception, Virtual Meeting Room and Virtual Auditorium, and per Call Routing Rule. Restrictions can also be applied across an entire deployment. Bandwidth limitations cannot be applied to the forwarding connection between a Proxying Edge Node and a Transcoding Conferencing Node. For more information, see Managing and restricting call bandwidth.