What would be a better solution?

[sfx2000]

Define what?

The Key Performance Indicators - and how to apply test cases that are consistent and repeatable. In the past, you've done great work there, but the Wireless Environment we see today is very different than it was 10 years ago when 802.11ac was in their first gen of devices.

The number of clients within a BSS/ESS have increased significantly, and it's not just PC's and Laptops - we've got mobiles and tablets, set top streamers, IoT devices like doorbells, cameras, sensors...

Every one of these device classes has a different traffic mix - so the older methods of let's put a Reference Client and the Router/AP under test and pushing a bunch of iperf streams doesn't reflect what happens in the real world.

This also should consider mixed mode operation - older IoT devices like printers might still be 11g, and we have to consider the mix between 11n, 11ac, 11ax, and soon 11be... very few WLAN's are homogeneous, rather they are a complex collection of devices with many different capabilities.

Speed is not the only metric, and capacity is a metric to be considered, but more importantly, how does the radio in the AP respond across different load levels and traffic patterns - and much of this would go into things like the MAC scheduler, and how QoS applies with that scheduler...

 

[Tech9]

the consumer version

Consumer version: AiMesh with Self-Defined Network, Game Boost and Open NAT.

Translation: Bunch of Repeaters with some VLAN support, QoS with data sharing to 3rd party company and Port Forwarding.

 

[sfx2000]

I've been running a mix of devices from 2-3 different brands and never had a problem with wireless handoff between them. One router, the other two as wired AP's over the past few years.

That should work fine in a conventional ESS - the AP's are just radios, and the DS between them is well defined in the specs...

 

[TheLostSwede]

That should work fine in a conventional ESS - the AP's are just radios, and the DS between them is well defined in the specs...

Just wanted to point out, since a lot of people are buying into the Mesh marketing, that WiFi works fine, even if you mix and match hardware and devices. Yes, there have been some client device issues (certain Intel and Broadcom compatibility issues), but in general, it works pretty well as a common standard, despite the best efforts of some companies to try and add proprietary features.

 

[Tech9]

WiFi works fine, even if you mix and match hardware and devices

My experience - works better when all the APs are the same type/specs and for consumer products the same model on the same firmware. Business class equipment with 3-4 APs on low power also works better than consumer class with 2 units at power limits. Similar to @tgl above I reused some older business equipment and still run two sets of Cisco RV345P + 4x Cisco WAP571. I may add some more updated firewall in front this summer.

 

[TheLostSwede]

My experience - works better when all the APs are the same type/specs and for consumer products the same model on the same firmware. Business class equipment with 3-4 APs on low power also works better than consumer class with 2 units at power limits. Similar to @tgl above I reused some older business equipment and still run two sets of Cisco RV345P + 4x Cisco WAP571. I may add some more updated firewall in front this summer.

More like when it's the same radios, i.e. Qualcomm, Broadcom or Mediatek, I have no real experience with Realtek routers.
I only had Qualcomm based hardware for a very long time. Then again, I had two Broadcom based Asus routers (one as an AP) and one old TP-Link router (also as an AP) and no issues roaming between the three still.

 

[tgl]

My experience - works better when all the APs are the same type/specs and for consumer products the same model on the same firmware.

One thing that I think you don't get with mix-n-match infrastructure is 802.11k/v roaming support from the APs. Now, client devices will certainly still roam from one AP to another without that, but the transitions might be a little glitchier without it. I've noticed slightly smoother roaming since I switched from a setup that didn't have 802.11k/v to one that does.

The reason you don't get this with unrelated APs is that at a minimum the APs have to trust each other as being okay roaming targets, and manufacturers seem to implement that as "the APs have to be configured from a common source". In some implementations the controller might be actively involved in 802.11k/v setup too; although with the UniFi APs I'm currently using, I know that the controller just tells them to do it, and the actual work of measuring neighboring APs' signal strength and so on is then done autonomously by the APs.

In ASUS-land I'd expect all the nodes in an AIMesh network to perform 802.11k/v signalling, whether they are exactly the same model or not. There are other good reasons behind @Tech9's recommendation that they be the same model, though.

 

[Tech9]

All identical APs eliminate one variable from preferred AP equation.

 

[sfx2000]

One thing that I think you don't get with mix-n-match infrastructure is 802.11k/v roaming support from the APs. Now, client devices will certainly still roam from one AP to another without that, but the transitions might be a little glitchier without it. I've noticed slightly smoother roaming since I switched from a setup that didn't have 802.11k/v to one that does.

802.11k/v/r have little to do with the radios specifically* - these are all infrastructure oriented specs, so using k/v/r does require core support across the DS...

* even if the radios support 11k/r - if the infra side, e.g. the DS, doesn't have support, it doesn't apply - remember that 11r/11v are more focused on enterprise deployments with radius on the backend - so for most home users using PSK, these options really do not apply...

One of the things I've observed with the QCA based Mesh implementations - there's a lot of reliance on the IEEE 1905 specs, and how QCA has implemented them within the QSDK board support packages for IPQ40xx and IPQ80xx and later chipsets.

 

[sfx2000]

More like when it's the same radios, i.e. Qualcomm, Broadcom or Mediatek, I have no real experience with Realtek routers.
I only had Qualcomm based hardware for a very long time. Then again, I had two Broadcom based Asus routers (one as an AP) and one old TP-Link router (also as an AP) and no issues roaming between the three still.

In a conventional ESS deployment - shouldn't matter what SoC is in use - heck, some vendors use a number of different chipsets and vendors - but the management layer they stay compatible...

QCA used to be dominant in the enterprise AP space, and they still have a significant share - what's interesting is seeing what's up with the CPE space - Quantenna was a big deal a couple of years ago with WiFi6 support, and suddenly they went away - replacements in my observations - about 50/50 with QCA and Broadcom for Cable DOCSIS, and most of this is split between Netgear and Vantiva (ex Technicolor).

5G-FWA - this is an interesting space - used to be QCA for 5G-WAN and the Wifi Radios, but MediaTek has made a lot of inroads there...

With the MediaTek based devices, with the right tools, one can force an OpenMesh deployment, even though it might not be supported... And the closed source mediatek firmware/drives - there's a lot of weird/clever stuff perhaps, not just openmesh, but other IBSS things like WiFi direct, etc

Only odd as I'm looking at this from a QCA perspective...

 

[TheLostSwede]

In a conventional ESS deployment - shouldn't matter what SoC is in use - heck, some vendors use a number of different chipsets and vendors - but the management layer they stay compatible...

QCA used to be dominant in the enterprise AP space, and they still have a significant share - what's interesting is seeing what's up with the CPE space - Quantenna was a big deal a couple of years ago with WiFi6 support, and suddenly they went away - replacements in my observations - about 50/50 with QCA and Broadcom for Cable DOCSIS, and most of this is split between Netgear and Vantiva (ex Technicolor).

Quantenna was apparently a PITA to integrate, as you still needed a third party SoC and Quantenna had a really weird interface early on, but even once that was improved, their drivers were apparently not that easy to integrate with the third party SoC. At Securifi we used a third party SoC, but they had the blessing of QCA, so it wasn't that hard to make it work with QCA WiFi. The same SoC was used by some of the big ISPs in the US as well.

5G-FWA - this is an interesting space - used to be QCA for 5G-WAN and the Wifi Radios, but MediaTek has made a lot of inroads there...

With the MediaTek based devices, with the right tools, one can force an OpenMesh deployment, even though it might not be supported... And the closed source mediatek firmware/drives - there's a lot of weird/clever stuff perhaps, not just openmesh, but other IBSS things like WiFi direct, etc

Only odd as I'm looking at this from a QCA perspective...

Qualcomm focused too much on the mobile device market and seemingly forgot they were making router hardware for the longest of time, so I believe they lost customers due to that, as well as higher pricing. MTK has obviously always been there at the entry level, alongside Realtek, which also seems to have more or less disappeared, despite showing WiFi 7 gear at Computex last year and presumably this year.
MTK seems to have opened up a bit more, at least that's what OpenWRT are suggesting and why they're going with MTK for their own developer hardware. I don't have the contacts I use to with these companies, so I can't really tell you what's going on in any detail.

 

[theoak]

Scooped up a TP-Link Deco BE85 (BE22000) WiFi 7 system during the Amazon Prime Day sales (3-Pack for $900) Amazon has been bouncing the prices on this kit all over the place since. After the sale it went back to $1500. They then dropped the 2 pack in price. They now have the 3 pack at $1100.

Anyway, I am using a router with 10 gig ports. The Decos are in access point mode. The BE85s have two 10 gig ports; thus, I have 10 gig wired backhaul. The Decos as per TP-Link are the only ones that allow aggregation over wired backhaul. I am thinking that is a WiFi 7 thing though.

https://www.tp-link.com/us/support/faq/3749/

Of course I had to get a few Intel BE200s for my laptops. Works great. Multi bands are selected. Depending on where I am in the home aggregated link speeds of over 5k.

It just works, I'm good.

Edit:

If you plan on keeping your Archer though, then stick with Archer. As per TP-Link (see link above), for optimal mesh, you should not mix Decos with Archers.

Also, for WiFi 7 and Windows 11, I had to get on the Dev release channel of Windows 11 24H2.

 

[sfx2000]

Quantenna was apparently a PITA to integrate, as you still needed a third party SoC and Quantenna had a really weird interface early on, but even once that was improved, their drivers were apparently not that easy to integrate with the third party SoC. At Securifi we used a third party SoC, but they had the blessing of QCA, so it wasn't that hard to make it work with QCA WiFi. The same SoC was used by some of the big ISPs in the US as well.

Quantenna was eventually acquired by OnSemi back in 2019 ($1B USD, what an exit!) and then wound it up and shut it down in 2022 - so that explains why we don't see any newer CPE with their silicon...

Their radios weren't back, but as you note, quite a few integration challenges...

Broadcom and QCA have picked up the slack in the Cable/DSL CPE market, and those are known quantities - over in 5G-FWA, it's QCA and MTK show there..

Qualcomm focused too much on the mobile device market and seemingly forgot they were making router hardware for the longest of time, so I believe they lost customers due to that, as well as higher pricing. MTK has obviously always been there at the entry level, alongside Realtek, which also seems to have more or less disappeared, despite showing WiFi 7 gear at Computex last year and presumably this year.
MTK seems to have opened up a bit more, at least that's what OpenWRT are suggesting and why they're going with MTK for their own developer hardware. I don't have the contacts I use to with these companies, so I can't really tell you what's going on in any detail.

QCA has been interesting to watch - between their automotive work, AI accelerators, and now PC chipsets, they're working to diversify into this new markets (for them) - in mobile, they're still pretty solid as they have been due to long term arrangements with the OEM/ODM's - WiFi, as you mention, it's an opportunity for them to improve...

They did well in the WiFi5 Wave 2 segment - WiFi6 they were competitive on the tech side, but perhaps not as many design wins - MTK has been super aggressive with WiFi 6/6E, and they do have the home field advantage with the ODM's over in Hsinchu Park...

MTK has made some interesting moves - with the mt76 support, they've done a couple of things - one is being a lot more open about getting changes into the mainline, and second is engaging with the OpenWRT community - a lot of the devs that were really engaged with WiFi like ath9k, they've moved over to mt76, so it's very interesting at the moment.