PeerTube server infrastructure

[and1]

Hi,
I’m currently researching and testing the best way to organize a PeerTube infrastructure.

We expect around 32,000 hours of video uploads per year, primarily uploaded in batches and later consulted. Given this workload, what would be the best approach?

1. Multiple servers, each hosting a separate PeerTube instance with dedicated storage and runners, distributing users across them.
2. A single instance with a large, fast attached storage and multiple runners handling the processing.

If we choose the second option, how is the load balanced across multiple runners?
Additionally, if users upload more (not yet transcoded) videos than the runners’ available storage, how does PeerTube handle this? Is there a queue or some kind of exception management?

Thank you!

[Chocobozzz]

Hi,

@and1

Given this workload, what would be the best approach?

It depends how many concurrent viewers you expect? See Tests de stress PeerTube : la résilience réside dans les pairs ! | JoinPeerTube

@and1

If we choose the second option, how is the load balanced across multiple runners?

Runners just pick up the first available job if they can.

@and1

Additionally, if users upload more (not yet transcoded) videos than the runners’ available storage, how does PeerTube handle this?

Files of pending jobs are kept on the peertube instance, and the peertube instance expects enough disk storage to store user uploads (that can be limited using user quota