Assuming zero load from any other services on the server, what number of simultaneous users on a single call/room could I reasonably expect to support before the service degraded?
I realise this is not something to be precise about, but a ballpark would be helpful. ~x10 ~x20 ~50x? users?
While the server specs are definitely good, the focus will need to be on the server bandwidth. That’s the simple most important resource for Jitsi/WebRTC.
Very roughly 1000/2.5=400. It means roughly receiving 200 HD streams and sending 200 HD streams. You can do one room with 200, 2 rooms with 100 each … and so on …
Note that you can probably lower the CPU cores and RAM a bit. Multiple JVBs of 4 or 8 cores with, say, 8 GB RAM work much better than one monster JVB, and JVB won’t use all that RAM anyway. With multiple servers you will probably get more bandwidth from your provider too.
VPS have a GigaBit vNIC - so bandwith is up to 1 Gbit/s
I wouldn’t assume that. If you’re not paying for a guaranteed 1 Gbit/s per server, you’ll likely hit shaping or congestion earlier, at least some of the time. Do some tests before your users test it for you