Scaling is critical for large deployments of BBB. The way of doing it is through a Load Balancer that estimates the demand and matches with the capacity a cluster may have defined.
Some LBs include the option for reserving certain capacity on create to avoid overloading the servers. But for that it is required to know what is the expected number of attendees in a meeting. Ideally, there should be a size limit per room or meeting, but when that is not established, a hint on the potential size of the audience would be the least data required.
The meta parameter used by LBs is `bbb-meeting-size-hint` and should be set to either the number of users enrolled in the course or group (when groups enabled).