Force adapter to connect to a specific WiFi band based on IP address?

[Grisu]

there will be a loss of about 3% per additional guest SSID for a specific band, not realy worth to think about a lot if you dont have more than 3 guest SSID per band.
I had this config too for some years.
main SSID same for 2/5G for family
one guest for friends same for 2/5G (sometimes only for 5G)
one for home automation only 2G
for each band and router seperated "test"SSIDs to test or use very band on certain router (routerA_2G, routerA_5G, routerB_2G, routerB_5G for all routers and APs).

 

[Smokindog]

What is the "one device" and what adapter does it have? Many adapters these days have settings for "preferred band" and other connection properties to help you control this situation.

Hi,

RT-AC86C 384.12_alpha1-g40c9e42009

Apologies if this has been covered, I looked but could not find a specific response.

I am using Smart Connect on this router but there is one device (and one only) that I wish to force onto the 5GHz band. I am aware I can define two separate SSIDs and connect FROM the device to the 5GHz one and I believe that I could tweak Smart Connect Rules (although it might sometimes connect and sometimes not?) but I really just want to force one WiFi adpater onto a specific band based on a simpler method.

I guess I could also use the Wireless MAC Filter for that one device's MAC with an accept / reject rule? I tried it but it did not work. For this method, if I put in a filter for "Reject"ing that MAC on the 2.4GHz band, will it then still connect on 5GHz, or will the MAC address filter apply to ANY subsequent Wifi connections and refuse all connections? Or do I ALSO need to Accept the 5GHz Wifi (with 2.4GHz rejected)?

Lastly, as I have the device (and 20-odd others from all the kids gizmos) reserved via DHCP to an IP Address anwyay, I was wondering of it is possible to connect to a defined band by IP Address only?

TIA,

k.

 

[Klueless]

It does mean the loss of guest access on one channel (as I do not wish them to “Access Intranet")

You should be fine. You can have up to three guest SSIDs for each radio and they can each use -or not use- intranet access as you see fit.

• All the tests resulted in the same download speeds except for Test 2 where the speed was about 7% slower ...

• ... So I am not sure what to make of all of this

In the "right" location the 86U can download up to 500 Mbps over the 5 GHz radio. Your Internet service of 300 Mbps could be the "pinch" point. In other words your WiFi speed would need to be below 300 before you could start calculating any degradation of service due to additional SSIDs.

Said differently; Your WiFi speed is 400 but your speed test is only going to show 300 because that's your Internet service cap. Add a few SSIDs, WiFi speed drops to 350 but you will still test out at 300 (because 350 is still greater than 300).

As far as test #2; it's WiFi. There's always going to be some randomness : -) You could watch your "traffic monitor" to make sure nothing else is going on at the same time. Then there's always the variability of the Internet (and your speed test partner) at any moment in time.

Might have better results if you ran your tests on your intranet rather than the Internet?

 

[Rumboogy]

You should be fine. You can have up to three guest SSIDs for each radio and they can each use -or not use- intranet access as you see fit.

In the "right" location the 86U can download up to 500 Mbps over the 5 GHz radio. Your Internet service of 300 Mbps could be the "pinch" point. In other words your WiFi speed would need to be below 300 before you could start calculating any degradation of service due to additional SSIDs.

Said differently; Your WiFi speed is 400 but your speed test is only going to show 300 because that's your Internet service cap. Add a few SSIDs, WiFi speed drops to 350 but you will still test out at 300 (because 350 is still greater than 300).

As far as test #2; it's WiFi. There's always going to be some randomness : -) You could watch your "traffic monitor" to make sure nothing else is going on at the same time. Then there's always the variability of the Internet (and your speed test partner) at any moment in time.

Might have better results if you ran your tests on your intranet rather than the Internet?

Yes exactly. At a ISP bandwidth of 300 Mbps I essentially saw no measurable degradation (except for the anomalous Test2). So what this tells me is that for the real world situation of my home environment it makes no difference if I use 1, 2, 3, or 4 total SSIDs for the 5 GHz. The degradation is below the noise of measurement. So this comment that it is "silly" to freely use guest SSIDs is just not true - at least under my test conditions.

Yes, I was using Traffic Manager to make sure the network was quite overall during testing. The "Test 2" test case was definitely odd. This was not a noise issue - I sampled it many times. Something was going on in that config.

Just for fun I am doing another test where I will only measure traffic on the WLAN. I will use a disk speed program on one machine connected via Ethernet writing to a RAM disk on another machine connected over 5 GHz. If that can't tease out the real effect of multiple SSIDs that I am out of ideas.

 

[Klueless]

Well, close enough for government work : -)

If you were to connect from a further corner of the house where you only get like 2 - 250 Mbps you could see degradation. But, like you (and the chart) say, a few is fine, but, a lot ... not.

I've been teased for running four and five but the reward (Value = Benefits - Cost) is worth it.

 

[jksmurf]

What is the "one device" and what adapter does it have? Many adapters these days have settings for "preferred band" and other connection properties to help you control this situation.

It’s a TP-Link USB WiFi Adapter T4U v2 that I have attached to a raspberry Pi2 running Libreelec Kodi 19 (Alpha); it’s drivers are out of tree (only recently learned what that means), but currently supported.

The v3 of that Adapter doesn’t run on Libreelec.

I also have a Netgear A6210 which is in the latest Kernel 5.1.3, is recognised by Libreelec and works initially but drops WiFi (separate issue).

I want to run HD playback (Pi doesn’t do 4K) and 5Ghz seems the faster connection.

There are settings in Windows for preferred band, granted; but nothing so easy (for me) in Linux/Libreelec.

k

 

[jksmurf]

there will be a loss of about 3% per additional guest SSID for a specific band, not realy worth to think about a lot if you dont have more than 3 guest SSID per band.
I had this config too for some years.
main SSID same for 2/5G for family
one guest for friends same for 2/5G (sometimes only for 5G)
one for home automation only 2G
for each band and router seperated "test"SSIDs to test or use very band on certain router (routerA_2G, routerA_5G, routerB_2G, routerB_5G for all routers and APs).

Cheers; good to know. With two routers I can probably have one SSID for the AC86U (family and guests) and for the A68U split one of the 4 off (actually two by default if you separate it).

 

[jksmurf]

You should be fine. You can have up to three guest SSIDs for each radio and they can each use -or not use- intranet access as you see fit.

You’re right; had forgotten the option to have 3 guest WiFi SSIDs; I cannot see an option for access to the intranet however, I understood the purpose of guest Wifi was that such access was restricted (by design)? [EDIT] OK, answered myself. This is only on the AP; which does not do any routing. It is available on the Router.

And with the tests just done by young Rumboogy, it seems like there will be no real issues for what I wish to do. Especially after having tried the MAC filter reject 2.4Ghz option which my adapter seems reluctant to want to “play nice” with.

 

[Rumboogy]

Using speedtest, even as you have (averaging) is not nearly consistent enough I would think. A better test would be to see a constant download from a wired desktop computer to a wireless one like your MBP, preferably using synthetic 'data' that doesn't rely on an HDD or other, possibly variable, data surface.

The results are still interesting, but as you hint at, not really worth using as 'proof' of the spreadsheets concepts.

I did a 2nd experiment to test the WLAN speed without relying on an ISP at all. The conditions were all the same as the first experiment (5 GHz radio, very little local traffic, same router & firmware, etc.) except:

• As the data source I used a Mac connected by gigabit Ethernet directly to the router.
  • The source Mac ran program by Blackmagic called Disk Speed Tool.
    
  • This was used to generate "synthetic" network traffic by writing to a mounted disk on the destination.
• For the destination I used another Mac connected to the router over 5 GHz WiFi (same as before).
  • On the destination Mac I setup a 6GB RAM disk.
    
  • I tested the write speed to the RAM disk form the destination Mac running the Blackmagic tool and got a write speed of of about 37 giga bits per second. That should be fast enough not to interfere with the relatively glacial speed of WiFi.
• I mounted the RAM disk across the network to the source Mac.
  • I then ran Blackmagic using a block size of 5 GB to write then read to the RAM disk over the network.
  • The data captured in the table below is write data (which is the download direction from the point of view of the destination Mac).
  The test cases are the same as before:
  • Test 1: main SSID only
  • Test 2: main SSID + guest 1
  • Test 3: main SSID + guest 1 + guest 2
  • Test 3: main SSID + guest 1 + guest 2 + guest 3

The results still show no slowdown when adding SSIDs (in some cases there was a slight speed up but this is probably due to noise). Here is all the data.

So the net-net is that using up to four SSIds (the max for ASUS) really does not mater at all - at least when on the 5 GHz radio in an congested area.

 

[Grisu]

it probably goes down a little bit if you - at the same time - really use those guest wifi with other clients connected to it with little traffic.
There are headers in wifi protocol and speed must go down somehow with more SSIDs.
Maybe you are in a range you wont realize it.
As I previoulsy said, upto 3 SSID per band nothing to think about at all!

 

[Klueless]

I did a 2nd experiment to test the WLAN speed without relying on an ISP at all ... So the net-net is that using up to four SSIds (the max for ASUS) really does not matter at all

I gotta admit I'm in awe. I wish I had half of your ambition!

(My grandkid's been complaining about performance/speed when he's playing "Fortnite" so I think I'll add a coupla SSIDs : -)

Just as an aside it looks like your overall WiFi speeds are pretty impressive as well. Somewhere between 480 Mbps (net speed) and 600-ish (because I have no idea how to account for the extra bits we used for packet overhead ... headers, trailers, etc.).

Just out of curiosity, what do you suppose the speed woulda been had you run Ethernet to Ethernet?

BTW: "net-net", that was kinda funny : -)

Again, I was really impressed!

 

[Klueless]

I did a 2nd experiment ...

@Rumboogy

So the 2nd test eliminated a couple of the variables I had teased you about (Internet variability and the 300 Mbps pinch point) but test #2 still yielded similar results; not only do we see minuscule impact of mulitple SSIDs but we actually see some improvement. Bizarre.

The way I understand it a beacon is sent for each SSID and each beacon is a data packet that takes space/bandwidth.

I looked at the spread sheet @L&LD was nice enough to share. It came preset to the rather draconian limits of an old 802.11 b implementation. As such it was claiming a whopping 3.22% per SSID.

• Just resetting that to "normal" B dropped that to just 0.43%

You tested on the 5 GHz band. Since the spread sheet hasn't been updated for N nor AC I selected the pop down for ancient A over 5 GHz and it displayed 0.12% overhead per SSID. Just for grins I selected G over 2.4 GHz and it came up with 0.11% overhead.

Now it's hard to believe that N and AC over 5 GHz would maintain backwards compatibility with A but, worse case, it still suggests that 4 SSIDs would use less than a half of a percent of bandwidth!

So there should be no significant difference.

But there was (and mostly for the better I might add). Tests within a test suite were consistent. It's when we ran a new test suite that things changed. What happened? Well for one we changed the router config which was almost like a genuine make-believe mini reset?

• Perhaps we got a different link speed after the change? Probably not.
  • @Rumboogy is demonstrating 500 Mbps and we saw the same anomaly when we had the 300 Mbps pinch point.
  • I tried similar and saw no change.
• Perhaps "beam forming" changed? Probably not for the same reasons as above.
• Perhaps there is some "consistent" randomness of the CA in CSMA/CA? E.g., the new timing is such that there are fewer "collisions" and retransmitts?
• Maybe it's got something to do with MU/MIMO? On the "MO" we've got router beacons, source Mac and destination Mac acknowledging. Maybe for one suite they all used the same stream, another suite shared only two streams and on another suite they all got different streams?

In other words I still have no klue.

So, with a couple caveats, I agree with you. If you need 'em use 'em!

 

[Klueless]

there will be a loss of about 3% per additional guest SSID for a specific band, not realy worth to think about a lot if you dont have more than 3 guest SSID per band.
I had this config too for some years.
main SSID same for 2/5G for family
one guest for friends same for 2/5G (sometimes only for 5G)
one for home automation only 2G
for each band and router seperated "test"SSIDs to test or use very band on certain router (routerA_2G, routerA_5G, routerB_2G, routerB_5G for all routers and APs).

Assuming "Router B" is on a different channel I count 4 SSIDs on your 2.4 GHz band. You are suggesting that 4 times about 3% comes to about 12% overhead.

When I tweaked the spread sheet I came up with more like 0.43% per SSID thus about 1.72% overhead for four SSIDs.

Thus (if I'm right and bear in mind I am "Klueless") you are more than right when you say, "Don't worry about it!"

 

[Grisu]

I just used this sheet for rough estimation years ago and came to the conclusion never mind upto 2-3 guest SSIDs: https://revolutionwifi.blogspot.com/p/ssid-overhead-calculator.html

 

[Klueless]

I just used this sheet for rough estimation years ago and came to the conclusion never mind upto 2-3 guest SSIDs: https://revolutionwifi.blogspot.com/p/ssid-overhead-calculator.html

And I'm saying you are probably even more right than you ever thought you were : -)

 

[Rumboogy]

I gotta admit I'm in awe. I wish I had half of your ambition!

(My grandkid's been complaining about performance/speed when he's playing "Fortnite". I think I'll add a coupla SSIDs : -)

Just as an aside it looks like your overall WiFi speeds are pretty impressive as well. Somewhere between 480 Mbps (net speed) and 600-ish (because I have no idea how to account for the extra bits we used for packet overhead ... headers, trailers, etc.).

Just out of curiosity, what do you suppose the speed woulda been had you run Ethernet to Ethernet?

BTW: "net-net", that was kinda funny : -)

Again, I was really impressed!

Thanks. The computer I did the testing on was 25-30 feet from the router with two walls in between. The Network Map View List shows is having around 2 bars on the 5 GHz. It is certainly faster than any ISP plan that I have had (or plan to get in the near future).

As far as your Ethernet to Ethernet question, I would assume it would go close to the 1 Gbps speed of the adapters. I guess it would be fairly easy to test but I only have on Mac with native Ethernet jack at this point. I am not sure how much using an Ethernet dongle would distort the measurement. In any event, gigabit Ethernet is so much faster than WiFi or my ISP that it just does not matter.

 

[Rumboogy]

@Rumboogy

So the 2nd test eliminated a couple of the variables I had teased you about (Internet variability and the 300 Mbps pinch point) but test #2 still yielded similar results; not only do we see minuscule impact of mulitple SSIDs but we actually see some improvement. Bizarre.

The way I understand it a beacon is sent for each SSID and each beacon is a data packet that takes space/bandwidth.

I looked at the spread sheet @L&LD was nice enough to share. It came preset to the rather draconian limits of 802.11 b. As such it was claiming a whopping 3.22% per SSID.

• Just resetting that to "normal" B dropped that to just 0.43%

You tested on the 5 GHz band. Since the spread sheet hasn't been updated for N nor AC I selected the pop down for ancient A over 5 GHz and it displayed 0.12% overhead per SSID. Just for grins I selected G over 2.4 GHz and it came up with 0.11% overhead.

Now it's hard to believe that N and AC over 5 GHz would maintain backwards compatibility with A but, worse case, it still suggests that 4 SSIDs would use less than a half of a percent of bandwidth!

So there should be no significant difference.

But there was (and mostly for the better I might add). Tests within a test suite were consistent. It's when we ran a new test suite that things changed. What happened? Well for one we changed the router config which was almost like a genuine make-believe mini reset?

• Perhaps we got a different link speed after the change? Probably not.
  • @Rumboogy is demonstrating 500 Mbps and we saw the same anomaly when we had the 300 Mbps pinch point.
  • I tried similar and saw no change.
• Perhaps "beam forming" changed? Probably not for the same reasons as above.
• Perhaps there is some "consistent" randomness of the CA in CSMA/CA? E.g., the new timing is such that there are fewer "collisions" and retransmitts?
• Maybe it's got something to do with MU/MIMO? On the "MO" we've got router beacons, source Mac and destination Mac acknowledging. Maybe for one suite they all used the same stream, another suite shared only two streams and on another suite they all got different streams?

In other words I still have no klue.

So, with a couple caveats, I agree with you. If you need 'em use 'em!

Definitely one should feel free to use the SSIDs as needed (at least on the 5 GHz radio when in low congestion).

I did not play around with the spreadsheet. I did not ever realized that I could - I saw it was locked and did not try further. Your analysis is interesting.

As far as what caused the noise in my experiments, who knows. One thing I should add though is that by the time I finished gathering all the data my router was definitely sick. As I got towards the end of gathering my experimental data, each time I turned an SSID on or off all the radios would go off and I had to reboot. So perhaps whatever this growing problem was impacted my data. To resolve the problem I had to: reloading firmware, factory resetting, loading JFFS & config from an earlier known stable point.

 

[Klueless]

... computer I did the testing on was 25-30 feet from the router with two walls in between ... around 2 bars on the 5 GHz

And you're getting > 500 Mbps ... that seems really good to me!

... router was definitely sick ... had to: reloading firmware, factory resetting, loading JFFS & config from an earlier known stable point

Ouch. Turned out to be a lot more work than you signed up for! But I wound up learning a lot from you. Hope your router is feeling better.

 

[Klueless]

Adding more SSID's is adding more overhead, which reduces throughput for all devices on the WLAN.

While it is true a lot of us do overstate the overhead as per this link:

https://revolutionwifi.blogspot.com/p/ssid-overhead-calculator.html

It's misleading as that page shows SSID overhead for an early implementation of 802.11 B over 1 Mbps. It shows a whopping 3.22% per SSID per AP on any given channel:

If you play with the spread sheet it shows more like 0.43% for the B over 11 Mbps that we are all accustomed to. Look at all the "green" now!

G over 54 Mbps shows only 0.11% overhead:

The spreadsheet was written before N and AC but if you look and A over 5 GHz it shows about 0.12%. One should assume N and AC only gets better?

SSIDs: It would seem like if you need 'em - use 'em. I also "think" it would be reasonable -in this day and age- to run N only on the 2.4 GHz band to further reduce overhead?

As long as we're talking about SSIDs I typically use something like this:

• Family - 2.4 & 5 GHz (family password)
• Faster - 5 GHz only (family password)
• Further - 2.4 GHz only (family password)
• Guest - 2.4 & 5 GHz, intranet disabled (guest password)

If anyone is having trouble they can typically figure out what to do.

 

[Klueless]

As far as your Ethernet to Ethernet question, I would assume it would go close to the 1 Gbps speed of the adapters

I only ask because sometimes I use Ethernet to Ethernet for my "calibration set" ... to see what the tool, workstations, NICs, etc. are capable of. E.g., I've a couple PCs that only get 250Mbps over Ethernet on a 400 Mbps Internet service so it's not really a surprise that they don't get the full 400 over WiFi either.