Throughput drop between switches

I have a NWA50AX with OpenWRT installed & configured on it. I put it on a TP-Link TL-SX1008 10Gb unmanaged switch, then connect a wireless client to it, load up fast.com, and get around 600 Mbps in throughput. Awesome.

I take the exact same access point and connect it to a TP-Link TL-SG1008MP 1Gb PoE+ unmanaged switch, then connect the same wireless client to it, load up fast.com, and get around 10 Mbps in throughput. Terrible!

What is going on here? I'd expect better numbers than this. I'm hoping this setup will replace my existing Unifi AC Pro (older unit) + Unifi Switch 16 w/PoE+--which is also a 1Gb switch--but that gets 210 Mbps!

I can connect a wired client to my TL-SG1008MP and get a full 1 Gbps from fast.com, so I know the switch can support those speeds. I'm quite puzzled why I can't get speeds faster than 10 Mbps on my OpenWRT AP on a 1 Gb switch. Would love any insight y'all have on the issue.

Sounds like this switch isn't working quite right... could be a number of things

First thing I would recommend is to disconnect the cable from the AP and plug that into a wired computer and see what you get for both actual throughput and the link speed that is shown on the physical interface.

Assuming it's still bad, disconnect the uplink cable from the switch and plug that cable into your computer. Do you still get the same speeds?

It could be a cable issue, bent pins or debris in the ports, or a problem with the switch itself. Start with the above, and we'll start to narrow in on the problem.

Did this and got 1 Gbps from the wired computer to the switch running a speedtest. The host OS (Ubuntu) reports the link speed as 1000 Mb/s.

Tried different cables & ports between the AP and the switch—same result.

FWIW the switch and the NWA50AX are both brand new.

How long is the cable between the switch and the AP?

Various lengths. Tried with a 6 ft cable and a longer run that's more on the order of 50 ft. Same performance.

I have another UniFi access point that's not used I'm going to try to setup a new wireless network on and test on the switch. I'm thinking that if that's good, it will go a long way to exonerate the switch itself as the problem.

If the switch checks out on the first test, I also have another NWA50AX (also running OpenWRT) that I'm planning on flashing with the default ZyXel firmware and try that. If that checks out, I feel like the problem must be the combination of this switch & OpenWRT... somehow...

Not likely... this is very likely physical in nature, and an unmanaged switch has no ability to interact in any strange ways with any hardware, unless you have some negotiation issues.

Another thought...
Keep your AP plugged into the switch and then connect a wired computer to another port on that switch. Does the wired computer still get good speeds? or does it experience degredation?

No performance degredation.

What does the AP report as its link speed? Is the AP running stock firmware or OpenWrt?

AP is running OpenWRT 23.05. Is the link speed displayed somewhere in LuCI? Or can I get it via some command over an SSH connection?

Port status is about 1/2 down the system status page in LuCI.

1 Like

LuCI reports 1 GbE on the connection.

But weirdly enough, now I'm getting 600+ Mbps on the AP & the TL-SG1008MP?? I swear nothing changed... I stepped away for a call, came back, and now it's revving at full throttle.

Gonna see if I can replicate the slower speeds now. Very weird. Not confidence inspiring as by not knowing what the original problem was I don't know how to fix it if it happens again...

Lets try to just test your network devices: Choose one of these and lets stress your network/intranet.

It could be a mechanical issue such as with the port or other similar things where the position or thermal state could cause it to have intermittent behavior. Hard to know without more testing/data.

That said, let's review your config to make sure there's nothing odd there:

Please connect to your OpenWrt device using ssh and copy the output of the following commands and post it here using the "Preformatted text </> " button:
grafik
Remember to redact passwords, MAC addresses and any public IP addresses you may have:

ubus call system board
cat /etc/config/network
cat /etc/config/wireless
cat /etc/config/dhcp
cat /etc/config/firewall

10 MBit/s is kind of a canary value, which suggests the switch being somehow running at 100 MBit/s, instead of 1 GBit/s. Are there any 100 MBit/s devices connected (printers, VDSL/ cable modems, etc.)? If there are, please consider unplugging them for a short test (and power cycle routers, switches, APs involved). While I don't expect any of this (it really shouldn't), it's an easy enough test to rule out something 'obvious' (it doesn't hurt to disconnect anything you don't strictly need for your test).

No 1000BASE-T switch should have an impact on your 802.11ax wireless at all, as you noticed -in practice- those tend to provide a maximum of 600-800 MBit/s for 2x2 clients, which is easy for any 1 GBit/s switch (unless there's lots of concurrent traffic on its uplink port).

1 Like

After some exhaustive testing, I've determined that my switch is faulty. The speeds would drop intermittently, which made verification hard, but in every combination, the performance degradation always occurred when the TL-SG1008MP was somewhere in the chain. Swapping it out entirely and testing all permutations with the UniFi switch instead never showed the same drop.

I've initiated a return and got a replacement unit, and so far have not experienced any issues, so :crossed_fingers:I'm good from here on out. I'll post back on this thread if I run into the same behavior with the replaced switch... dealing with another issue with my ISP at the moment relating to IPv6, so I need to get that sorted first.

Thanks for all the help & feedback!

Glad you got it all sorted out! Based on the behavior you were experiencing, it really does make sense that it was a faulty switch

That was my initial reaction...

And you said:

And this was probably both annoying to you and the thing that led you to believe it was some other situation... but bad devices do escape factory testing, or fail for some other reason. Good to have resolution!

If your problem is solved, please consider marking this topic as [Solved]. See How to mark a topic as [Solved] for a short how-to.
Thanks! :slight_smile:

Yeah, super frustrating. I probably was guilty of blaming the more "complicated" component (OpenWRT) over the "simpler" component of a brand-new, unmanaged switch. The inconsistent behavior and my unfamiliarity with OpenWRT in general probably didn't help things much.

Thanks so much for your help & patience!

1 Like

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.