A low rssi means higher transmission power, higher chance of interference and hidden/exposed terminal problems, etc. ...
BUT, for me the approach of just disconnecting clients, seems a bit aggressive. Did someone made some measurements or made experience with that?
For me that is some kind of trade-off between disconnecting clients and wireless network performance. If you kick clients in that way, the concerning users of those devices will suffer a lot?
Also lower speed for those hosts with low RSSI as well as for everyone since it will take more time to send or receive something to or from them.
I made some field tests in my house and set the kick-RSSI as low as when the client device was walking outside, so yeah I would not need the Wifi any more.
Not more than they would suffer if they disconnected because of no-signal-at-all.
The analogy I use is to imagine a meeting of a group in a large room, where someone way in the back has to shout words out slowly and repeat them several times in order to be understood by the chairman in the front.
The overall situation would go much more smoothly to ignore him.
This approach only works (to some extent, and yes, this is debatable) in a roaming network (ideally with 802.11k/v/r) with many APs, to fight sticky clients and 'encourage' them to (ideally seamlessly) roam to another, stronger, AP (the whole concept relies on there actually always being a stronger AP in range, it fails towards the edges of this roaming domain).
Actually, I tried to implement client steering / load balancing in OpenWrt. [0]
You could actually do that with the daemon. If someone is interested, I can give you an appropriate configuration. Still this is under heavy development and I would not call that stable!
BUT, I noticed if you are trying to interfere to aggressive. Clients will have a bad time...
And in some home scenario aggressive kicking clients, without knowing if there is a better AP seems like making the WiFi network unusable for some clients that are to far away. And I doubt the performance increase of the network, if it only consists of maybe 3 WiFi stations...
Therefore, I'm interested how many stations I need that kicking makes sense. Or which configuration (ht, vht, 5 GHz)?
Ahhhh, the daemon can only disassociate clients, if they do background scanning. If they do background scanning, they typically will roam to a better AP themselves. So actually, the daemon is not able to do that.
You are right that in a home with 3 users this is not much of a concern at all. In a public situation with dozens of users that are constantly moving (and many will be within whatever you have defined as the "fringe" coverage), it is a lot more complicated.
Yep, even in a home situation, I have 3 APs, they cover house, front and back yards, and with a private and a guest network. If I can get the clients to switch reliably as they move from say front yard to back yard, they perform a lot better. A lot of clients want to stick to their tiny signal they initially get from the front AP. Booting them off makes them work a lot better when they're in the backyard. In my case by raising the lowest basic_rates.
I think that clients are sticky is some myth.
They try to save as much energy as possible but if they really want to transmit something or the rssi is to low they switch to other APs.
Could you give me some information about those clients? Old Android devices? Did u disable background scanning?
Yeah, things have definitely gotten better with time. My Moto G3 was particularly sticky, as were several other Android devices. Also a linux laptop with 2.4GHz only. My Moto G5 is much more fluid.
I think devices that are 2.4GHz only are often stickier.
