Question Strange Internet Performance Issues Over Wireless

[EightySix Four]

I am experiencing the strangest internet performance issues that I have ever encountered and could use some additional ideas for how to test and narrow down the problem. The setup:

• 1gbps symmetric internet connection
• 802.11ax router
• One wired linux server which provides DNS for the network
• One wired Xbox Series X
• Multiple wireless devices with 802.11ax support including a second desktop with an Intel AX201, two M1 Macs (one Pro, one Air), and an iPhone 13 Pro

The wired devices pull near lgigabit connections (900+ mbps) to speed test servers (speedtest.net, fast.com, etc.) while the wireless devices cannot download at more than 350mbps or so, usually trending towards high 200mbps. Initially I assumed it was the obvious - the signal wasn't good enough and the connection is being downgraded to G/N or the 2.4ghz networks, so I tried to find a better channel/channel width combination, but no matter what I did, I could not break through 350mbps or so, even when my adapters were reporting >1000mbps connections.

So, I thought I would test the wireless connection by setting up an OpenSpeedTest docker container on my wired linux server. Surprisingly, my wireless devices were able to connect to my wired server at near gigabit speeds as well, usually 850mbps or so.

What I have tested up until this point:

• WAN -> WLAN is slow
• WAN -> LAN is fast
• LAN -> WLAN is fast

Then I decided to setup a SOCKS5 proxy on my linux server and try and run speedtest from one of the wireless devices through the proxy and it was at nearly gigabit speeds.

So now we have:

• WAN -> WLAN is slow
• WAN -> LAN is fast
• LAN -> WLAN is fast
• WAN -> LAN -> WLAN is fast

At this point I reach out to router vendor number 1, they escalate me to R&D who sets up teamviewer sessions, confirms the behavior I am seeing, and takes a bunch of logs from my device. Then they sent me a different model router which has a different OS and processor that is also 802.11ax and I get the same results.

Today, annoyed at the wireless issues and having been through multiple long troubleshooting session with vendor #1, I purchased a router from a competing vendor and, much to my surprise, got the same results. So now I am here very confused about where else this could be an issue in my network. Is it possible my ISP is throttling wireless connections somehow? Have I stumbled upon some hardware issue that both wireless vendors have with a shared chipset or something?

 

[Tech Junky]

the signal wasn't good enough and the connection is being downgraded to G/N or the 2.4ghz

You already know the answer......

adapters were reporting >1000mbps connections.

expect 50%+ of the LR reported

I have a LR of 2.4gbps and max out at ~1.5gbps internally on the LAN / ~1200mbps on the WAN depending on the VPN server I'm connected to.

my wireless devices were able to connect to my wired server at near gigabit speeds

Now this is interesting since you should expect about the same performance across the router backplane.

Knowing how much a POS and bloatware these cheap routers have on them and your testing would point to some sort of DPI / AV / QOS mess inspecting packets from WLAN for some odd reason. Disable all of the gimmicks on the router and see what happens when testing again w/o the S5 in the middle hiding the content.

BTW I built my own router using a Linux PC and hanging an AP off of it for these reasons of TSing vendor "features".

 

[EightySix Four]

All three routers ha all of the “fun” Firewall/AV/DPI/QoS stuff turned off, but even if they were the cause, they should be causing it on the wired connections as well.

I am willing to believe even two products from the same vendor have a problem, but I am struggling with the idea that three products from two vendors, all running different OSes have the same problem.

Also worth noting that these issues occur with and without IPv6, which my ISP supports natively.

 

[Tech Junky]

So now we have:

• WAN -> WLAN is slow (except w/ SOCKS5)
• WAN -> LAN is fast
• LAN -> WLAN is fast
• WAN -> LAN (SOCKS5)-> WLAN is fast

3 devices / 3 WIFI cards / 2 OS's

Wrapping the packets / changing the headers shouldn't really make a drastic difference IMO.

WAN <cable> Modem <cable> Router / WIFI

All ancillary "fun" disabled so, it shouldn't be a CPU issue on the router

This sort of thing I would do a packet trace on from the router to see where the packets are hanging but, that's a Cisco thing. It's odd that funneling everything through a PC would result in higher throughput. Then again that's exactly what I've been doing for a few years now w/ a Linux PC as the router / firewall / AP (internal & external) for the pure speed of the setup.

The only thing I can think of that might be happening is the SOCKS server is pre-pending a higher classification to the packets forcing them to be prioritized on the router as something like voice or video but, if QOS is disabled in the router and at gig speeds isn't really necessary anyway unless you have a ton of clients.

From my server I can ping my laptop and average a 2ms ping. From the laptop to the server it does bounce around a bit more 1-3ms. So, there's a bit of jitter in play but, not enough to be noticeable w/o looking at the pings. Right now though I'm using a Zyxel AP w/ a 2.5GE uplink to the server through a POE injector vs the internal PCI card setup I was before with AC.

 

[EightySix Four]

Just tested with a VPN, it also allows the traffic to go faster.

Is my ISP identifying wireless traffic and throttling it? How?

 

[Tech Junky]

What router are you using for WIFI or are you using an AP attached to a router?

If you're using the ISP router for WIFI I suppose it's possible that they might deprioritize the WIFI traffic that's not masked within a proxy or VPN. It seems more like DPI on the WIFI traffic though which doesn't really make sense as it's usually performed on ALL traffic.

It doesn't make sense to me. Routed traffic through SOCKS aka PC <> wired = fast / VPN cloaked WIFI = fast. There has to be something in the router / AP that's looking at WIFI traffic for some reason that's not apparent.

When you're running the VPN connection I'm assuming you're doing it per client and not on the router.

 

[EightySix Four]

I’m using my own router, I’m actually not sure about the modem hardware, I live in a condo building and have Ethernet directly into my unit which is connected to fiber (probably in a networking closet somewhere).

Correct on the VPN connection, but doing it at the router seems like an excellent test!

 

[Tech Junky]

Well, if you have a fiber drop to ONT / Ethernet then it's a higher end device somewhere in the building and shouldn't be looking at wired / wifi traffic differently.

What kind of router are you using? Might lead to figuring out if there's a bug or HW issue.

As I said earlier I VPN everything outbound for that extra layer of security / anonymity from everyone good and bad. wireguard gives me full speeds compared to OVPN which chops everything to less than 50% of the capable speed due to the way it works.

 

[EightySix Four]

The issue was duplicated on a TP-Link Archer AX6000, Archer AX90, and an Asus RT-AX86U. The two TP-Links are on different chipsets and OSes.

 

[fkoehler]

So why not remove all the variables but one?

Disconnect LAN from WAN, and connect 1 wirelss router to the WAN, and connect a known good PC to the wifi and test.

I'm not seeing how your ISP is going to know what connections are from the LAN vs WLAN, and thus be able to throttle respectively.

IME, when you run across problems like these that seem to be not following the laws of reality, its usually come down to some piece of equipment either not doing what it is supposed to according to IEEE standards, or it IS doing what its supposed to, but not documented properly.

Interesting problem.

 

[RearAdmiral]

Looking forward to seeing how this gets resolved. It seems like the issue is upstream from all of your stuff....somehow. Does anyone else in the condo have the same service that you could run tests with?

 

[EightySix Four]

I have some time to do some testing this weekend. One of my tests is going to be setting the router into AP mode and connecting a single device to it so it is basically a wireless bridge directly to WAN.