Hello,
Here is my setup
Fiber Modem (bridge) > OpenWrt main router ( PPPoE + VLAN) > Dumb AP
When I connect a PC to the main router via Wired and do a speed test SQM work fine <1ms
Whenever any device behind the Dumb AP does a speed test the ping spikes ( +20-50ms )
What's missing?
Appreciate your efforts, insights, criticism and everything in between.
Most likely the problem here is the wifi which gets congested much easier than the cable connection directly to the router.
Is the connection between the OpenWrt router and the dumb AP wired or wireless and what sort of WiFi connection do you have to your dumb AP? And what sort of transfer rates are we dealing with?
I have even just a wireless connection between my OpenWrt router and dumb AP and at my sub-100Mbit/s rates cannot see any noticeable WiFi-induced delay (less than 5ms on top of my circa 45ms LTE connection).
@dtaht, @moeller0 and @amteza have lots of experience in this area. Any advice gents?
it's definitely not most likely, this haven't come to mind even a bit. the DUMP ap is tplink AX1800 which theoretically can handle 1.2Gbps on the 5ghz network with no issue even though my internet speed is only 300Mbps, when set the SQM for 100 there is no ping spikes 0-5ms , when raise it to 200 the ping increased a little 5-10ms, with 300 it become +40ms.
I would appreciate any suggestion for the replacement to fix the ping spikes.
What the iperf3 rate from client over WiFi to your OpenWrt router? You can test by e.g. running iperf3 -s on your OpenWrt router and then iperf3 -c 192.168.1.1 from client.
The theoretical speeds are marketing bloated numbers for optimal conditions which don't apply in real life.
Other than that, make sure no other devices use the access point, the channel is not congested from neighbornets, and that the client is using the 5G band without obstacles.
Is this however an issue or just an experiment that didn't go as planned? I mean, do you often download at maximum speed and worry about the latency?
My internet plan was 100Mbps which was fine to handle by Qos of the tplink, the provider offered a free upgrade to 300Mbps and most router can't handle Qos of such a speed.
C:\iper>iperf3.exe -c 192.168.2.1
Connecting to host 192.168.2.1, port 5201
[ 4] local 192.168.2.185 port 51368 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-1.00 sec 25.4 MBytes 213 Mbits/sec
[ 4] 1.00-2.00 sec 26.6 MBytes 223 Mbits/sec
[ 4] 2.00-3.01 sec 25.0 MBytes 209 Mbits/sec
[ 4] 3.01-4.00 sec 28.0 MBytes 236 Mbits/sec
[ 4] 4.00-5.00 sec 28.1 MBytes 236 Mbits/sec
[ 4] 5.00-6.00 sec 26.9 MBytes 225 Mbits/sec
[ 4] 6.00-7.00 sec 29.6 MBytes 248 Mbits/sec
[ 4] 7.00-8.00 sec 29.4 MBytes 247 Mbits/sec
[ 4] 8.00-9.00 sec 28.0 MBytes 234 Mbits/sec
[ 4] 9.00-10.00 sec 29.2 MBytes 245 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-10.00 sec 276 MBytes 232 Mbits/sec sender
[ 4] 0.00-10.00 sec 276 MBytes 232 Mbits/sec receiver
iperf Done.
Even though the WiFi show link speed 400 Mbps on Windows
WiFi link speed is a gross rate, actually usable is considerable less, often in the 50-66% range (you milage will vary...)
Have you tested that actually wih locally sourced and sinked traffic?
Mmmh, this means either your router runs out of CPU between 200 and 300 Mbps, or your achievable WiFi rate through the dumb AP simply is < 300 Mbps on average.
Uses ethernet cables or use SQM on the dump AP and shape the WiFi well below 200 Mbps (or what ever shaping rate that device can actually sustain).
That bandwidth seems to me rather low.
It's even less than on my WDS-setup in which I have dumb AP's connected wirelessly to my main OpenWrt router - 3x RT3200's. I see circa 750Mbit/s between OpenWrt and dumb AP and circa 350Mbit/s between client and OpenWrt router indirectly via dumb AP.
Then again, this is in the middle of the woods in the Scottish Highlands. So no interference whatsoever.
I would first try and work out if you can improve the WiFi bandwidth. Are you dealing with lots of interference and if so have you optimised as much as possible channel, bandwidth and AP placement?
Let's start with the fake 1.2 Gbps number, and to be very clear it does not matter what the vendor says. Your connected device is likely the one limiting you. Let's say it's a common device 2x2 MIMO (MCS 9, doubt you can get MCS 11), you will be talking about 960 Mbps which 60% (if you are lucky and your air is not congested in the channel) will be your real bandwidth (one way, as WiFi is not full-duplex). So, there you go 576 Mbps.
Now, if your bandwidth with the ISP is 300 Mbps, you should configure your SQM to be between 85-95% of it, let say 90%. To see the real benefit of SQM you must use 270 Mbps as the bandwidth configured in SQM.
QQ, is your AX1800 running OpenWrt? If not, expect spikes of +20 ms at least under heavy load. For a real world example see the following:
Test ran while downloading a file in another computer (computer connection 2x2 MIMO MCS 9 867 Mbps), FYI.
https://www.waveform.com/tools/bufferbloat?test-id=9f6095d9-0924-45c0-84ca-ed3ea72de14c
In summary, TL;DR, what you are seeing when you don't limit your bandwidth (SQM set to 300 Mbps) is the poor dump AP latency in action.
Actually that's pretty good! 400 Mbps * 60% = 240 Mbps. You are using all the available bandwidth from that wireless link during the testing. 
Real world matters, mate.
@amteza My ISP provides 1gbit symmetrical Internet and what I set is 900mbit or even 850mbit and I never have such result on my AX3200 as you have.
Could you tell me, if you have enabled:
If all is enabled, could you share configuration?
Thanks
No hardware offload, no software offload. Packet steering enabled, though.
But, just to insist, I'm using a RBPi4 as my router, and not the AP by itself.
This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.
