开云体育

Hi Rick,

I agree they work with non-UniFi switches.? When multiple AP's broadcast the same SSID at the same?site, do devices roam across them well, or does roaming break things -- is the question I'm raising.

"MAC flapping" is the phenomenon I'm referring to, if we're looking to validate this as something beyond a theory I made up on my own.? Basically I'm referring to whichever way the switch behaves when it sees the same MAC address come across two different ports in a short timeframe, which is what happens in a multi-AP-same-SSID situation as users roam from AP to AP.? You are probably already familiar with the idea that switches build an internal MAC table in their memory to help them determine which port(s) to forward traffic to, and then there's nuances to how that works (e.g. the special mac-to-IP mappings for multicast, and then protocols like IGMP that straddle layer2 and layer3 to help the switch build that table, for IP-related multicasts like basic mDNS, etc.) and also that most of us aren't engineers and haven't WireSharked our way deep into the mechanics of what these packets look like.? But bottom line, the switch tracks which MAC addresses go to which physical ports (and also what multicast groups these devices have broadcast themselves as being a member of), and if the mappings must change (due to roaming) in a way the switch doesn't properly track, then some traffic is going to get misdelivered, with the same symptoms of having the connection lost.

You are probably also familiar with the fact that newer WiFi stacks, especially on mobile, will do some level of advanced forward negotiation like 802.11r when they're considering switching AP's, so they can seamlessly (for example) carry phone calls across the roam transition with minimal interruption / rather than drop the call for a second or two while they negotiate connection with a new AP.? Depending on what the potential new AP sends upstream to the switch during a transition, some of that forward negotiation may not play well the layer-2 implementation of an upstream switch not designed to deal with it, resulting in misdelivery of packets.? If you've got more insight, I'd find the gist of it interesting.

There's probably indeed some RFC or IEEE standard out there that defines exactly how modern switches should handle this,?and compliance with standards (rather than brand of switch) may be the reason for my reported success.? I'll maintain that I've had UniFi AP performance on Apple iOS devices go from continuously horrible (constant frozen WiFi connections every day, that seemed to mostly plague Apple devices) to perfect, with the single act of changing out the switch.? Each person's mileage may vary.

Mike

Contact me in PV

Regards,

Sébastien LENOIR

I once had pervasive reproducible issues like this on Apple products simply disappear overnight by switching the UniFi AP's from a non-UniFi to a UniFi switch.? Now I will never connect UniFi AP's to a non-UniFi switch, just not worth the hassle.? The best I can tell, these newer seamless fast-roaming protocols have more complexity than before, and if the upstream switch is behind on knowing which AP is actually in touch with an iPhone, and is sending traffic to the wrong AP for any reason, you'll drop packets and connections.? I assume Ubiquiti got the roaming dance figured out within their product line.

Mike