Am I right in my understanding that sharing the load only happens for people viewing the same video?
Could there be a benefit to P2P distribution of bandwidth that isn't just at the video level, but perhaps at the instance level, where they choose to federate with other instances that also agree to share the P2P load? Or rather, to share the P2P wealth?
That's the p2n layer that already exists in PeerTube through the redundancy feature in federation.
Federated instances that enable redundancy copy all resolutions of a video that meet chosen criteria of popularity. When a user views the video on PeerTube, their browser grabs the sections of the video from any of the PeerTube instances that have a copy, creating an ad-hoc distributed CDN. This covers the mid-level of popularity and keeps the hosting instance from getting slammed on bandwidth.
If something goes viral, or during livestreams which are by nature simultaneous, then the p2p in the browser between viewers kicks in and reduces load on the peering servers even more.
That's great! And along the lines of what I was hoping might be true. A question though: It sounds like you are saying various instances can help each other on a per-video basis, for videos above a popularity threshold.
But what about the prospects of distributing the load that isn't localized to a particular video? As in, users do some peer to peer load distribution of any and all videos above the threshold, regardless of whether it's the one they are watching?
Could there be a benefit to P2P distribution of bandwidth that isn't just at the video level, but perhaps at the instance level, where they choose to federate with other instances that also agree to share the P2P load? Or rather, to share the P2P wealth?