Connection flood issue

[Popolechien]

Not too sure where this should be posted, and this is more a behavioural report that a solvable issue, but @kelson42 suggested this should be documented in any case.

Was at Wikimania with a demo hotspot (Preppers Premium H1). The room had 2-300 attendees and the hotel wifi was struggling to connect everyone (connections would drop every so often). I plugged the RPi and even for that one connectivity was very patchy / slow (and it's not like it had been around for long (like, 30s), so the odds of 200 people seeing it and trying to connect were low).

[rgaudin]

OK, good to know but indeed we can't do much about it because there are not enough technical details, we can't reproduce easily and the details you gave highlight a scenario that we don't care about.

It's really a bad situation because in this case tons of people are frequently being disconnected and thus scan the network. Then there's the fact that our network is open so depending on the device, failing to connect to the Hotel one might lead them to automatically connect to the open one (or manually for those looking at it… which probability increases with 300 people).

I suppose the wlan chipset was overwhelmed. There are other possibilities of course: the kernel, due to the routing we impose for the captive portal (doubt it), the captive portal also is not optimized for numerous parallel request (possibly creating a feedback loop on the connections). Might also have been on same channel as the Hotel…

While I think this is not a problem for us to solve, being able to demo to a large size room is important. That would require dedicated WiFI hardware carefully configured, documentation, etc. A mini project of its own.