Dear forum members,

The topic of this post is frequently dealt with on the Internet, but it does not provide me the right information, so I will specify more precisely my question in this forum.

In a research project we want to provide wireless access to a large number of users. The use case for this research project is not to provide plain Internet access, but to provide event information to the visitors of the event using a closed wireless network. For this purpose we build an app that visitors install on their mobile and which controls what a user can do. Data transfer consists of video distribution towards the users, and additional low bandwidth data exchange with an event server.

Could you tell me what the bottlenecks are in the maximum number of (simultaneous) users per access point, and whether these bottlenecks stem from architectural choices, implementation choices or are just scalable with the processing power of the computing platform. I know the number of associations is limited to 2007, but what other limitations are there when setting up the system using COTS HW, the OpenWRT platform and hostapd? And how can these bottlenecks be avoided or mitigated? Also notice that per user bandwidth is not so much an issue because of multicast video distribution and telemetry-like unicast traffic.

Looking forward to your response.

Regards, Jan

Hostapd has a default limit of 150, but that can be overridden. 

This sort of thing is very difficult to test.  The "Make Wifi Fast" and "Bufferbloat" projects attempt to optimize in the face of heavy use, but I think they mostly concentrated on situations of tens or fewer users using heavy bandwidth each.

In a large festival situation you could have a lot of users who connect then move away from the AP and still try to use it with a weak signal. This is very bad on performance.  I have a project which may be of interest there, see github.com/mk248.

(Last edited by mk24 on 18 Jan 2018, 15:01)