How do Chime clusters work?
Chime intelligently groups participants' browsers together to form clusters. By establishing clusters, a meeting will save on bandwidth and conference resources by offloading some of the central collaboration capability. A cluster is typically comprised of endpoints in close network proximity to one another that transmit audio and video directly between themselves (peer-to-peer). Network proximity is not based on physical or geographic proximity but rather network performance, as measured by delay and jitter metrics between the endpoints being monitored by the Chime server.
Chime dynamically controls bit rate, resolution and frame rate to minimize the encoding and decoding burden across participants. With the active speaker, Chime uses larger bitrates, resolutions and frame rate, but non-active participants use less bandwidth and smaller resolutions. As more parties are added to a conference, we can be judicious in video resolution & quality.
Clusters can be created and destroyed as necessary over the life of a meeting. If network conditions change, participants and uplink nodes can be removed from a cluster at any time and reconnect through the Chime server directly. For customers that choose not to use clustering, it can be disabled as a configuration option.