Hi I've heard the general consensus that Wifi roaming is purely done by the client devices instead of the mesh satellites / WAPs. However here under "The problem with non-mesh access points" and "The advantages of mesh networking" PCWorld claims that mesh networks have smarts built into them to facilitate better roaming on the wireless network. Is this true/untrue? Maybe half-true? Thanks!
-
AMD deep-dives Zen 5 architecture — Ryzen 9000 and AI 300 benchmarks, RDNA 3.5 GPU, XDNA 2, and more
Question Mesh Network Smarts?
- Thread starter Crag_Hack
- Start date
First why would you need so called seamless roaming. I will laugh when router manufactured start getting sued because someone fell down the stairs in their house staring at netflix on their phone. They already walk out into traffic on cell phones.
So many many years ago cisco tried this and most the information about the 802.11 roaming protocols you will find on cisco sites. This was back in the day when you paid by the minute for cell calls and coverage in buildings was horrible. Cisco tried to do what is now called wifi calling but it had a different name back then. It kinda worked but not well. In the end the cell company killed it off with unlimited voice plans and the use of micro cell tower placed in large buildings.
As stated the client is in full control and it does not pick the correct AP. Now the part they try to pretend are the roaming extensions to the 802.11 protocols. Most end devices do not support this so it really doesn't matter if the mesh units do. Again the client is still in control. Even when is has the roaming protocols it is just a bit more information to the end device. The huge problem is the signal level the AP detect from the end device do not mean the end device gets the same signal levels from the AP. The mesh network has no way to determine how well the end device can receive signals from any AP. It makes the assumption it is symmetrical which many times is correct but also why you still get random disconnects and drops as you move around.
Something like a cell phone the tower is in full control of the radio in your cell phone. It can adjust the transmit power on the fly and tell the radio to change towers.
BUT we are again back to why do you really need this feature. In most cases if you walk down the hall by the time you get to the other room the device will detect it and switch even using just dumb AP with no mesh. Nobody in a home environment is going to try to dynamically balance devices. The example cisco used was a bunch of employees move to a conference room and all want to use wifi. The system would force other users who were not on the conference room to move to other AP to allow the extra capacity. Again it kinda worked and kinda didn't
So many many years ago cisco tried this and most the information about the 802.11 roaming protocols you will find on cisco sites. This was back in the day when you paid by the minute for cell calls and coverage in buildings was horrible. Cisco tried to do what is now called wifi calling but it had a different name back then. It kinda worked but not well. In the end the cell company killed it off with unlimited voice plans and the use of micro cell tower placed in large buildings.
As stated the client is in full control and it does not pick the correct AP. Now the part they try to pretend are the roaming extensions to the 802.11 protocols. Most end devices do not support this so it really doesn't matter if the mesh units do. Again the client is still in control. Even when is has the roaming protocols it is just a bit more information to the end device. The huge problem is the signal level the AP detect from the end device do not mean the end device gets the same signal levels from the AP. The mesh network has no way to determine how well the end device can receive signals from any AP. It makes the assumption it is symmetrical which many times is correct but also why you still get random disconnects and drops as you move around.
Something like a cell phone the tower is in full control of the radio in your cell phone. It can adjust the transmit power on the fly and tell the radio to change towers.
BUT we are again back to why do you really need this feature. In most cases if you walk down the hall by the time you get to the other room the device will detect it and switch even using just dumb AP with no mesh. Nobody in a home environment is going to try to dynamically balance devices. The example cisco used was a bunch of employees move to a conference room and all want to use wifi. The system would force other users who were not on the conference room to move to other AP to allow the extra capacity. Again it kinda worked and kinda didn't
K thanks for the excellent answer. Sounds like the PCWorld article claims including "Disruptive interruptions to connection quality, especially with real-time applications like streaming and video conferencing." resulting from non-mesh networks shortcomings are for the most part bogus.
If you were talking about online games where even adding a extra 200ms will cause a lag spike maybe. I want to see someone actually walking around there house carrying their pc and trying to play a shooter type game.
Most stuff like streaming and video conferencing have buffers that add delays of multiple full seconds in many cases. This tends to hide most this type of issue. You can sometime detect the delays kinda like when you watch nasa talk to astronaughts
I can see very little real world use case. In most cases if you are using internet based stuff you might as well just used a cell network that is designed from the ground up for seamless roaming.....you can still hear the static as it switches cell tower when you are driving so even that is not truly seamless. In your house you would never change towers as you walk around.
The only application I can see that needs very high bandwidth and needs some mobility is VR. There are already enough video of people hurting themselves falling over furniture in the same room. Are they really dumb enough to try to walk down the stairs with a vr headset on.
This is mostly marketing hype. People have learned the scam that all brands of routers pretty much have the same coverage. It isn't really the coverage that is the problem it is that the coverage is actually too large and all your neighbors signals are coming into your house. So you must sit closer to the router to get the same amount of usable bandwidth from your router. To increase coverage they now use multiple radio units in everyone's house which just makes the problem worse and worse.
There is so much mesh stuff being sold that some vendors now want people to think they actually "need?" seamless roaming.
Most stuff like streaming and video conferencing have buffers that add delays of multiple full seconds in many cases. This tends to hide most this type of issue. You can sometime detect the delays kinda like when you watch nasa talk to astronaughts
I can see very little real world use case. In most cases if you are using internet based stuff you might as well just used a cell network that is designed from the ground up for seamless roaming.....you can still hear the static as it switches cell tower when you are driving so even that is not truly seamless. In your house you would never change towers as you walk around.
The only application I can see that needs very high bandwidth and needs some mobility is VR. There are already enough video of people hurting themselves falling over furniture in the same room. Are they really dumb enough to try to walk down the stairs with a vr headset on.
This is mostly marketing hype. People have learned the scam that all brands of routers pretty much have the same coverage. It isn't really the coverage that is the problem it is that the coverage is actually too large and all your neighbors signals are coming into your house. So you must sit closer to the router to get the same amount of usable bandwidth from your router. To increase coverage they now use multiple radio units in everyone's house which just makes the problem worse and worse.
There is so much mesh stuff being sold that some vendors now want people to think they actually "need?" seamless roaming.
- May 29, 2013
- 37,732
- 3,435
- 156,790
There is almost no way to predict. Besides interference, there are walls. The construction materials make a big difference. A lot of European homes have cement or concrete block interior walls. WIFI doesn't penetrate concrete.
"minimize overlap through using different channels" 
They are talking out both sides of their mouth when they talk about channels.
The so called channels are 5mhz wide on 2.4g radio. In theory there are 11 or 13 depending on country. So 60mhz total.
The minimum actual channel width used by wifi is 20mhz. It is much more common to use 40mhz to get higher speed. This means it is impossible to fit 2 signals without overlap.
On the 5ghz band the channel numbers do represent 20mhz bands but they now use 40,80 and even 160mhz of bandwidth. The 160mhz ones will not fit without overlapping things like weather radar.
We then get the 6ghz band which should solve this problem there is massive amounts of bandwidth but the bandwidth pigs are back.
Wifi7 now is going to use 320mhz channels. To make things even worse it will attempt to use the maximum radio bandwidth it can get on 2.4, 5 & 6 all at the same time, attempting to bond it all into 1 big pipe. This just 1 router not even discussing mesh or multiple units. Even if you were to play nice one of your neighbors is going to try the new bonded wifi7 routers and devices when they finally get onto the market.
They are talking out both sides of their mouth when they talk about channels.
The so called channels are 5mhz wide on 2.4g radio. In theory there are 11 or 13 depending on country. So 60mhz total.
The minimum actual channel width used by wifi is 20mhz. It is much more common to use 40mhz to get higher speed. This means it is impossible to fit 2 signals without overlap.
On the 5ghz band the channel numbers do represent 20mhz bands but they now use 40,80 and even 160mhz of bandwidth. The 160mhz ones will not fit without overlapping things like weather radar.
We then get the 6ghz band which should solve this problem there is massive amounts of bandwidth but the bandwidth pigs are back.
Wifi7 now is going to use 320mhz channels. To make things even worse it will attempt to use the maximum radio bandwidth it can get on 2.4, 5 & 6 all at the same time, attempting to bond it all into 1 big pipe. This just 1 router not even discussing mesh or multiple units. Even if you were to play nice one of your neighbors is going to try the new bonded wifi7 routers and devices when they finally get onto the market.
BFG-9000
Splendid
- Sep 17, 2016
- 3,542
- 500
- 29,590
Cellular towers communicate with each other so can seamlessly handoff your phone to the next tower when the time is right, or even block your phone entirely if it is seen by hundreds of towers at once (if you leave it on in an airplane).
Wifi is not like that, and clients often will refuse to move to an AP with a stronger signal until the signal to the one it's connected to drops entirely. 802.11r fast transition reduces the time to connect to a different AP to reduce the interruption (it's still not seamless) by allowing the key to be cached, which is of course what's targeted by KRACK.
Theoretically, if the APs/nodes communicate with each other they could decide only the one with the strongest signal should connect and all the others should kick the device off (similar to how the band-steering feature works, where 5GHz-capable clients are identified and get periodically kicked from 2.4GHz), but I'm not aware of any MESH setup that actually does this. Instead MESH is usually just sold as an expensive but convenient way to administer all of the APs at once using one GUI screen, and with a wireless backhaul to avoid having to run cable to APs.
It's usually enough to just adjust the roaming sensitivity of your clients so they will switch easier, and in the APs to reduce the Roaming Assistant setting until they occasionally kick a client with lower signal, so that client will then reconnect to whatever is strongest then. Yes, this will still incur a slight delay when switching, long-distance low-signal connections will be repeatedly disconnected, and switching necessarily won't be that frequent because in order to force the client to do so, an AP has no option other than to kick the client off to force a renegotiation.
As for overlapping channels, we are used to thinking in terms of 20MHz-wide blocks (so will manually set 2.4GHz to the non-overlapping ch1, 6 and 11) but OFDMA in Wifi 6 and above can automatically divide things even more to 10MHz or even 5MHz wide "resource units" as needed, perfect for many low-bandwidth but latency-sensitive streams. Selecting different channels than your neighbors helps minimize co-channel interference, but they may not be aware each 2.4GHz channel is only 5MHz wide so for example choosing a 20MHz-wide block centered on ch4 would interfere with channels 2, 3, 5 and 6, leaving only a 20MHz block centered on ch9-11 as non-overlapping for their neighbors
With 5GHz people usually choose either 80MHz-wide ch36-48 or ch149-161, or even 160MHz-wide using both non-contiguously (80+80 mode) just to avoid the long connection delay inherent in using DFS channels.
Historically 802.11b was spread-spectrum which helped to reduce the effects of interference, since packets were widely spread over different frequencies, leaving empty space over the air (so with luck interference could happen to just fit in the unused parts). 802.11a on 5GHz and 802.11g on 2.4GHz improved bandwidth by basically using every bit of their 20MHz-wide single pipes with no empty space, and this meant of course that any interference anywhere in that pipe would result in corrupted packets and a retransmit. It's been a solid pipe ever since, but at least 320MHz and 160MHz can either fall back to unbonded, or even use OFDMA to fall back to even 5MHz-wide if needed due to interference.
Wifi is not like that, and clients often will refuse to move to an AP with a stronger signal until the signal to the one it's connected to drops entirely. 802.11r fast transition reduces the time to connect to a different AP to reduce the interruption (it's still not seamless) by allowing the key to be cached, which is of course what's targeted by KRACK.
Theoretically, if the APs/nodes communicate with each other they could decide only the one with the strongest signal should connect and all the others should kick the device off (similar to how the band-steering feature works, where 5GHz-capable clients are identified and get periodically kicked from 2.4GHz), but I'm not aware of any MESH setup that actually does this. Instead MESH is usually just sold as an expensive but convenient way to administer all of the APs at once using one GUI screen, and with a wireless backhaul to avoid having to run cable to APs.
It's usually enough to just adjust the roaming sensitivity of your clients so they will switch easier, and in the APs to reduce the Roaming Assistant setting until they occasionally kick a client with lower signal, so that client will then reconnect to whatever is strongest then. Yes, this will still incur a slight delay when switching, long-distance low-signal connections will be repeatedly disconnected, and switching necessarily won't be that frequent because in order to force the client to do so, an AP has no option other than to kick the client off to force a renegotiation.
As for overlapping channels, we are used to thinking in terms of 20MHz-wide blocks (so will manually set 2.4GHz to the non-overlapping ch1, 6 and 11) but OFDMA in Wifi 6 and above can automatically divide things even more to 10MHz or even 5MHz wide "resource units" as needed, perfect for many low-bandwidth but latency-sensitive streams. Selecting different channels than your neighbors helps minimize co-channel interference, but they may not be aware each 2.4GHz channel is only 5MHz wide so for example choosing a 20MHz-wide block centered on ch4 would interfere with channels 2, 3, 5 and 6, leaving only a 20MHz block centered on ch9-11 as non-overlapping for their neighbors
With 5GHz people usually choose either 80MHz-wide ch36-48 or ch149-161, or even 160MHz-wide using both non-contiguously (80+80 mode) just to avoid the long connection delay inherent in using DFS channels.
Historically 802.11b was spread-spectrum which helped to reduce the effects of interference, since packets were widely spread over different frequencies, leaving empty space over the air (so with luck interference could happen to just fit in the unused parts). 802.11a on 5GHz and 802.11g on 2.4GHz improved bandwidth by basically using every bit of their 20MHz-wide single pipes with no empty space, and this meant of course that any interference anywhere in that pipe would result in corrupted packets and a retransmit. It's been a solid pipe ever since, but at least 320MHz and 160MHz can either fall back to unbonded, or even use OFDMA to fall back to even 5MHz-wide if needed due to interference.
TRENDING THREADS
-
How to fix CrowdStrike BSODs in three minutes — fix requires manual changes, but they are simple- Started by Admin
- Replies: 31
-
-
Question Dell AMD 5900x doesn't work in other motherboards...???- Started by TonyVier
- Replies: 4
-
Facebook
Twitter
Instagram