FYI/A: Bufferbloat 101

xiaofan · Sep 15, 2024

Ubuntu Linux 24.04 test results
1) Singtel 5Gbps wired connection with 5Gbe NIC --> A

Singtel ONR 10G LAN port --> Hasivo 2.5G Switch with 10G links --> Hasivo 10G SFP+ switch with RJ45 copper module --> Ethernet cable --> Ryzen 5 5600G desktop (IOcrest 5Gbe NIC)

https://www.waveform.com/tools/bufferbloat?test-id=5c0b8c3c-92c8-45ec-8b7d-cd0e061c10fd

2) Singtel 5Gbps wired connection with 10Gbe NIC --> A+

Singtel ONR 10G LAN port --> Hasivo 2.5G Switch with 10G links --> Hasivo 10G SFP+ switch with RJ45 copper module --> Ethernet cable --> Ryzen 5 5600G desktop (Intel X520-DA2 SPF+ with RJ45 copper module).

https://www.waveform.com/tools/bufferbloat?test-id=8677efec-8a1c-4c2d-a1e1-c72cd9a888d6

xiaofan · Sep 15, 2024

John Martino said:
which exact post? with wireless?

anyway here's mine with MyRepublic (1gbps) ONT -> nanopi R4SE with Cake/pieceofcake SQM 800down/700up -> TP Link Deco x20 -> Wired to work laptop (Intel NIC)

Using Windows client, results on Page 22 of this thread.
Post #437 --> wireless test, with and without sqm on OpenWRT side
Post #438 --> wired test, with and without sqm on OpenWRT side
https://forums.hardwarezone.com.sg/threads/fyi-a-bufferbloat-101.6427979/page-22

Then I have wireless test without SQM on OpenWRT side but with QoS on the ZTE BE7200 Pro+ AP side in this page. Same Windows client.
Post #442

xiaofan · Oct 3, 2024

@hwzlite

Now we have a very fast 4Gbps capable Crusader server for Singapore (hosted on Linode) thanks to the famous Dave Täht.

I have done my first wired tests with the server with my Intel N100 mini PC (Singtel 5Gbps plan, ZTE F8648P ONR) and the results are actually very good for both IPv4 and IPv6.
https://github.com/Zoxc/crusader/issues/9

Bash:

mcuee@miniroute10g:~/build/networking/crusader_bin/v0.3.1$ ./crusader test --load-duration 60 --streams 8 --stream-stagger 4 singapore.starlink.taht.net
[2024-10-03 12:48:51] Client version 0.3.1 running
[2024-10-03 12:48:51] Connected to server 172.236.148.60:35481
[2024-10-03 12:48:53] Idle latency to server 2.73 ms
[2024-10-03 12:48:55] Testing download...
[2024-10-03 12:50:25] Testing upload...
[2024-10-03 12:51:55] Testing both download and upload...
[2024-10-03 12:53:33] Warning: Load termination timed out. There may be residual untracked traffic in the background.

-- Download test --
          Throughput: 3927.62 Mbps
             Latency: 2.4 ms (0.8 ms down, 1.7 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 4935.37 Mbps
             Latency: 9.9 ms (0.9 ms down, 9.0 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 8427.17 Mbps (3579.31 Mbps down, 4847.85 Mbps up)
             Latency: 9.7 ms (1.0 ms down, 8.7 ms up)
         Packet loss: 0% down, 0.02% up

[2024-10-03 12:53:33] Writing data...
[2024-10-03 12:53:33] Saved raw data as crusader-results/crusader-results/test 2024-10-03 12.53.33.crr
[2024-10-03 12:53:33] Saved plot as crusader-results/test 2024-10-03 12.53.33.png

mcuee@miniroute10g:~/build/networking/crusader_bin/v0.3.1$ ./crusader test --load-duration 60 --streams 8 --stream-stagger 4 2600:3c15::f03c:95ff:fe7e:75a2
[2024-10-03 12:57:07] Client version 0.3.1 running
[2024-10-03 12:57:07] Connected to server [2600:3c15::f03c:95ff:fe7e:75a2]:35481
[2024-10-03 12:57:09] Idle latency to server 2.82 ms
[2024-10-03 12:57:11] Testing download...
[2024-10-03 12:58:46] Testing upload...
[2024-10-03 13:00:16] Testing both download and upload...
[2024-10-03 13:01:50] Warning: Load termination timed out. There may be residual untracked traffic in the background.

-- Download test --
          Throughput: 4057.47 Mbps
             Latency: 2.8 ms (1.0 ms down, 1.8 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 4873.10 Mbps
             Latency: 10.3 ms (0.1 ms down, 10.2 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 8280.88 Mbps (3577.16 Mbps down, 4703.72 Mbps up)
             Latency: 10.2 ms (0.6 ms down, 9.6 ms up)
         Packet loss: 0%

[2024-10-03 13:01:50] Writing data...
[2024-10-03 13:01:50] Saved raw data as crusader-results/crusader-results/test 2024-10-03 13.01.50.crr
[2024-10-03 13:01:51] Saved plot as crusader-results/test 2024-10-03 13.01.50.png

xiaofan · Nov 16, 2024

For those who are interested in latency testing (wired and wireless), you can take a look at here. Contributions of testing results are also welcome.
https://github.com/Zoxc/crusader/discussions

xiaofan · Mar 9, 2025

Coming back to this thread to see how wireless settings in the router can affect Bufferbload or Latency under load.

Test conditions
a) Fibre Internet: Singtel 5Gbps plan

b) Singtel issued ZTE F8648P ONR 10G LAN port bridged -->

virtual OpenWRT main router running on Intel N100 mini PC with Intel 82599ES 10G SFP+ port and no-brand SFP+ to 10G RJ45 copper module on the WAN side, and no-brand SFP+ to 10G optic module on the LAN side

--> fibre cable --> Hasivo 8-port 10G SFP+ switch with SFP+ 10G optic modules --> fibre cable

--> Hasivo 2.5G/10G switch with one SFP+ port with 10G SFP+ optic modue, and one 10G RJ45 port, and quad 2.5G ports

--> Asus RT-AX86U (AP Mode, always ON) and Asus TUF-BE6500 (router mode, Double NAT, ON for this test) and TP-Link Archer BE805 (router mode, Double NAT, not ON in this test).

[Do not ask me why I have such a complicated setup -- I will try to simplify it when I got the time to take down the network. One of the problem is somehow DAC cable did not work in my case, maybe a Hasivo SFP+ switch problem].

c) Wireless router -- Asus TUF-BE6500, with MLO ON and Smart Connect ON on the main SSID (2.4GHz+5GHz MLO). The IoT SSID will not have MLO feature as it is for legacy and IoT clients (2.4GHz/5GHz).

d) Wireless client -- Acer Swift Go 14 2024 model with Intel BE200 based Killer BE1750 WiFi 7 adapter.

e) Distance from the client to the router -- 3m

Test using waveform.com --> kind of useless for Singtel users.

1) Main SSID (2.4GHz+5GHz MLO) -- Grade "A".
The result does not make real sense though as the upload speed is very low.
https://www.waveform.com/tools/bufferbloat?test-id=efc6a7a0-bb3d-410f-8c4a-20e352d025f9

2) IOT SSID (5GHz) -- Grade "C" because of upload latency.
https://www.waveform.com/tools/bufferbloat?test-id=d4c504c4-7031-461a-ae2e-202be8e88bcf

xiaofan · Mar 9, 2025

OOkla SpeedTest results in the same setup.

No real differences between the main SSID (2.4GHz+5GHz MLO, WiFi 7) and IoT SSID (5GHz band, WiFi 6), in terms of speed and latency.

1) Main SSID (2.4GHz+5GHz MLO) -- latency is okay for both download and upload. Speed is also normal.

Bash:

PS C:\work\speedtest\ookla-speedtest-1.2.0-win64> .\speedtest.exe -s 13623

   Speedtest by Ookla

      Server: Singtel - Singapore (id: 13623)
         ISP: Singtel Fibre
Idle Latency:     4.69 ms   (jitter: 1.17ms, low: 4.45ms, high: 6.82ms)
    Download:  1461.14 Mbps (data used: 2.2 GB)
                 14.09 ms   (jitter: 8.44ms, low: 2.85ms, high: 272.41ms)
      Upload:  1059.54 Mbps (data used: 928.0 MB)
                 30.86 ms   (jitter: 13.57ms, low: 3.81ms, high: 106.45ms)
 Packet Loss:     0.0%
  Result URL: https://www.speedtest.net/result/c/ff066d40-5f42-4650-aa27-eb00cccbc5b8

2) IoT SSID (5GHz)

Bash:

PS C:\work\speedtest\ookla-speedtest-1.2.0-win64> .\speedtest.exe -s 13623

   Speedtest by Ookla

      Server: Singtel - Singapore (id: 13623)
         ISP: Singtel Fibre
Idle Latency:     3.51 ms   (jitter: 0.16ms, low: 3.21ms, high: 3.65ms)
    Download:  1432.91 Mbps (data used: 2.2 GB)
                 15.24 ms   (jitter: 8.76ms, low: 3.09ms, high: 269.44ms)
      Upload:  1129.60 Mbps (data used: 2.0 GB)
                 18.01 ms   (jitter: 7.58ms, low: 4.31ms, high: 80.76ms)
 Packet Loss:     0.0%
  Result URL: https://www.speedtest.net/result/c/e04cf055-1e45-4c70-a4d2-2469533f2ba1

xiaofan · Mar 9, 2025

More tests to be done with Crusader
https://github.com/Zoxc/crusader

Same test conditions as before.

Crusader seems to be a good tool for the latency test under load. It shows two potential issues.
a) Packet loss during download --> this may be related to my OpenWRT virtual router or the SFP+ to 10G RJ45 copper module (10G WAN port side) and 10G SFP+ optic module (10G LAN port side)
b) MLO SSID at close distance is not that good --> higher upload latency than the IoT SSID.

1) Main SSID (2.4GHz+5GHz, WiFi 7).

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 singapore.starlink.taht.net
[2025-03-09 20:39:12] Client version 0.3.2 running
[2025-03-09 20:39:12] Connected to server [2600:3c15::f03c:95ff:fe7e:75a2]:35481
[2025-03-09 20:39:14] Idle latency to server 5.26 ms
[2025-03-09 20:39:16] Testing download...
[2025-03-09 20:40:46] Testing upload...
[2025-03-09 20:42:17] Testing both download and upload...
[2025-03-09 20:43:51] Warning: Load termination timed out. There may be residual untracked traffic in the background.

-- Download test --
          Throughput: 901.29 Mbps
             Latency: 12.5 ms (6.1 ms down, 6.5 ms up)
         Packet loss: 0.48% down, 0% up

-- Upload test --
          Throughput: 905.65 Mbps
             Latency: 191.5 ms (65.0 ms down, 126.5 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 1320.31 Mbps (539.28 Mbps down, 781.04 Mbps up)
             Latency: 47.7 ms (11.8 ms down, 35.9 ms up)
         Packet loss: 0.04% down, 0% up

[2025-03-09 20:43:51] Writing data...
[2025-03-09 20:43:51] Saved raw data as crusader-results\test 2025-03-09 20.43.51.crr
[2025-03-09 20:43:52] Saved plot as crusader-results\test 2025-03-09 20.43.51.png

2) IoT SSID (5GHz, WiFi 6)

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 singapore.starlink.taht.net
[2025-03-09 19:55:56] Client version 0.3.2 running
[2025-03-09 19:55:56] Connected to server [2600:3c15::f03c:95ff:fe7e:75a2]:35481
[2025-03-09 19:55:58] Idle latency to server 6.08 ms
[2025-03-09 19:56:00] Testing download...
[2025-03-09 19:57:30] Testing upload...
[2025-03-09 19:59:06] Testing both download and upload...
[2025-03-09 20:00:45] Warning: Load termination timed out. There may be residual untracked traffic in the background.

-- Download test --
          Throughput: 775.71 Mbps
             Latency: 16.6 ms (5.6 ms down, 11.0 ms up)
         Packet loss: 0.06% down, 0% up

-- Upload test --
          Throughput: 923.58 Mbps
             Latency: 63.0 ms (10.2 ms down, 52.8 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 1207.26 Mbps (560.46 Mbps down, 646.80 Mbps up)
             Latency: 69.7 ms (19.9 ms down, 49.9 ms up)
         Packet loss: 0.06% down, 0% up

[2025-03-09 20:00:45] Writing data...
[2025-03-09 20:00:45] Saved raw data as crusader-results\test 2025-03-09 20.00.45.crr
[2025-03-09 20:00:45] Saved plot as crusader-results\test 2025-03-09 20.00.45.png

Edit to add the following to debug the packet loss issue.
1) using crusader server on the OpenWRT virtual router to remove the WAN side factor --> still see the packet loss

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 192.168.18.1
[2025-03-09 21:09:31] Client version 0.3.2 running
[2025-03-09 21:09:31] Connected to server 192.168.18.1:35481
[2025-03-09 21:09:33] Idle latency to server 2.74 ms
[2025-03-09 21:09:35] Testing download...
[2025-03-09 21:11:05] Testing upload...
[2025-03-09 21:12:35] Testing both download and upload...
[2025-03-09 21:14:09] Warning: Load termination timed out. There may be residual untracked traffic in the background.

-- Download test --
          Throughput: 975.92 Mbps
             Latency: 20.1 ms (7.7 ms down, 12.4 ms up)
         Packet loss: 0.47% down, 0% up

-- Upload test --
          Throughput: 1275.52 Mbps
             Latency: 109.3 ms (55.4 ms down, 53.9 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 1562.80 Mbps (427.25 Mbps down, 1135.56 Mbps up)
             Latency: 125.2 ms (12.5 ms down, 112.7 ms up)
         Packet loss: 0.06% down, 0% up

[2025-03-09 21:14:09] Writing data...
[2025-03-09 21:14:09] Saved raw data as crusader-results\test 2025-03-09 21.14.09.crr
[2025-03-09 21:14:09] Saved plot as crusader-results\test 2025-03-09 21.14.09.png

2) using another wired client of the TUF-BE6500 (Linux lxc container running on another Intel N100 mini PC) as the Crusader server to remove the main OpenWRT virtual router out of the picture. In this case, NAT is also out of the picture and the Crusader server is at the same local network segment (192.168.50.x) as the Crusader client.

No more packet loss. Upload test speed is now much faster but then latency increases a lot as well. So we can see wireless link can have greater latency under heavy load condition.

The conclusion of the packet loss issue: there is a potential problem with the main OpenWRT virtual router -- Intel N100 mini PC with dual 10G SFP+ ports (Intel 82599ES chipset).

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 192.168.50.15
[2025-03-09 21:44:32] Client version 0.3.2 running
[2025-03-09 21:44:33] Connected to server 192.168.50.15:35481
[2025-03-09 21:44:34] Idle latency to server 2.64 ms
[2025-03-09 21:44:36] Testing download...
[2025-03-09 21:46:06] Testing upload...
[2025-03-09 21:47:36] Testing both download and upload...
[2025-03-09 21:49:11] Warning: Load termination timed out. There may be residual untracked traffic in the background.

-- Download test --
          Throughput: 1785.92 Mbps
             Latency: 67.1 ms (59.5 ms down, 7.6 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 1400.95 Mbps
             Latency: 336.1 ms (10.3 ms down, 325.8 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 1796.67 Mbps (923.75 Mbps down, 872.92 Mbps up)
             Latency: 383.9 ms (11.2 ms down, 372.8 ms up)
         Packet loss: 0%

[2025-03-09 21:49:11] Writing data...
[2025-03-09 21:49:11] Saved raw data as crusader-results\test 2025-03-09 21.49.11.crr
[2025-03-09 21:49:11] Saved plot as crusader-results\test 2025-03-09 21.49.11.png

xiaofan · Mar 9, 2025

Basic wireless settings

5GHz band wireless professional settings.

xiaofan · Mar 9, 2025

More tests to be done to see if "Airtime Fairness" setting change anything or not. In the above tests, the setting is with default value "OFF".

Other potential setting changes to explore -- MU-MIMO and DL/UP OFDMA.

Then more testing to be done with far distance where MLO may be useful.

xiaofan · Mar 9, 2025

Testing conditions:

a) Wireless router -- Asus TUF-BE6500, with MLO ON and Smart Connect ON on the main SSID (2.4GHz+5GHz MLO). The IoT SSID will not have MLO feature as it is for legacy and IoT clients (2.4GHz/5GHz).

b) Wireless client -- Acer Swift Go 14 2024 model with Intel BE200 based Killer BE1750 WiFi 7 adapter.

c) Distance from the client to the router -- 3m

d) using another wired client of the TUF-BE6500 (Linux lxc container running on another Intel N100 mini PC) as the Crusader server to remove the main OpenWRT virtual router out of the picture. In this case, NAT is also out of the picture and the Crusader server is at the same local network segment (192.168.50.x) as the Crusader client.

Conclusion --> "Airtime Fairness" does not seem to help. Maybe the implementation by Asus in TUF-BE6500 is not good.

1. "Airtime Fairness" disabled (fefault)

1) MLO SSID (2.4GHz+5GHz MLO)

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 192.168.50.15
[2025-03-10 20:36:01] Client version 0.3.2 running
[2025-03-10 20:36:01] Connected to server 192.168.50.15:35481
[2025-03-10 20:36:02] Idle latency to server 2.78 ms
[2025-03-10 20:36:04] Testing download...
[2025-03-10 20:37:35] Testing upload...
[2025-03-10 20:39:05] Testing both download and upload...

-- Download test --
          Throughput: 1731.89 Mbps
             Latency: 75.5 ms (61.9 ms down, 13.6 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 1269.73 Mbps
             Latency: 316.0 ms (32.8 ms down, 283.2 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 1389.38 Mbps (736.15 Mbps down, 653.23 Mbps up)
             Latency: 708.7 ms (6.2 ms down, 702.6 ms up)
         Packet loss: 0%

[2025-03-10 20:40:39] Writing data...
[2025-03-10 20:40:39] Saved raw data as crusader-results\test 2025-03-10 20.40.39.crr

2) IoT SSID (5GHz)

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 192.168.50.15
[2025-03-10 21:20:13] Client version 0.3.2 running
[2025-03-10 21:20:13] Connected to server 192.168.50.15:35481
[2025-03-10 21:20:14] Idle latency to server 2.94 ms
[2025-03-10 21:20:16] Testing download...
[2025-03-10 21:21:47] Testing upload...
[2025-03-10 21:23:17] Testing both download and upload...

-- Download test --
          Throughput: 168.02 Mbps
             Latency: 93.6 ms (71.3 ms down, 22.4 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 85.53 Mbps
             Latency: 88.4 ms (3.7 ms down, 84.7 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 1412.43 Mbps (658.49 Mbps down, 753.94 Mbps up)
             Latency: 546.5 ms (16.6 ms down, 529.9 ms up)
         Packet loss: 0%

[2025-03-10 21:24:51] Writing data...
[2025-03-10 21:24:51] Saved raw data as crusader-results\test 2025-03-10 21.24.51.crr
[2025-03-10 21:24:51] Saved plot as crusader-results\test 2025-03-10 21.24.51.png

2. "Airtime Fairness" enabled

1) MLO SSID (2.4GHz+5GHz MLO)

This does not seem to help (tested twice).

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 192.168.50.15
[2025-03-10 20:47:03] Client version 0.3.2 running
[2025-03-10 20:47:03] Connected to server 192.168.50.15:35481
[2025-03-10 20:47:04] Idle latency to server 2.65 ms
[2025-03-10 20:47:06] Testing download...
[2025-03-10 20:48:36] Testing upload...
[2025-03-10 20:50:07] Testing both download and upload...

-- Download test --
          Throughput: 1909.17 Mbps
             Latency: 60.5 ms (49.2 ms down, 11.3 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 1133.47 Mbps
             Latency: 483.6 ms (44.2 ms down, 439.5 ms up)
         Packet loss: 0% down, 0.02% up

-- Bidirectional test --
          Throughput: 1520.15 Mbps (865.40 Mbps down, 654.76 Mbps up)
             Latency: 807.0 ms (13.0 ms down, 794.0 ms up)
         Packet loss: 0%

[2025-03-10 20:51:41] Writing data...
[2025-03-10 20:51:41] Saved raw data as crusader-results\test 2025-03-10 20.51.41.crr
[2025-03-10 20:51:41] Saved plot as crusader-results\test 2025-03-10 20.51.41.png

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 192.168.50.15
[2025-03-10 20:53:40] Client version 0.3.2 running
[2025-03-10 20:53:40] Connected to server 192.168.50.15:35481
[2025-03-10 20:53:42] Idle latency to server 2.79 ms
[2025-03-10 20:53:44] Testing download...
[2025-03-10 20:55:14] Testing upload...
[2025-03-10 20:56:44] Testing both download and upload...

-- Download test --
          Throughput: 1907.27 Mbps
             Latency: 61.0 ms (50.0 ms down, 11.0 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 1220.70 Mbps
             Latency: 152.8 ms (8.6 ms down, 144.2 ms up)
         Packet loss: 0%

-- Bidirectional test --
          Throughput: 1635.54 Mbps (959.33 Mbps down, 676.22 Mbps up)
             Latency: 887.2 ms (457.1 ms down, 430.1 ms up)
         Packet loss: 0% down, 0.01% up

[2025-03-10 20:58:19] Writing data...
[2025-03-10 20:58:19] Saved raw data as crusader-results\test 2025-03-10 20.58.19.crr
[2025-03-10 20:58:19] Saved plot as crusader-results\test 2025-03-10 20.58.19.png

2) IoT SSID (5GHz)

Bash:

PS C:\work\speedtest\crusader_bin\v0.3.2> .\crusader.exe test --load-duration 60 --streams 8 --stream-stagger 4 192.168.50.15
[2025-03-10 20:47:03] Client version 0.3.2 running
[2025-03-10 20:47:03] Connected to server 192.168.50.15:35481
[2025-03-10 20:47:04] Idle latency to server 2.65 ms
[2025-03-10 20:47:06] Testing download...
[2025-03-10 20:48:36] Testing upload...
[2025-03-10 20:50:07] Testing both download and upload...

-- Download test --
          Throughput: 1909.17 Mbps
             Latency: 60.5 ms (49.2 ms down, 11.3 ms up)
         Packet loss: 0%

-- Upload test --
          Throughput: 1133.47 Mbps
             Latency: 483.6 ms (44.2 ms down, 439.5 ms up)
         Packet loss: 0% down, 0.02% up

-- Bidirectional test --
          Throughput: 1520.15 Mbps (865.40 Mbps down, 654.76 Mbps up)
             Latency: 807.0 ms (13.0 ms down, 794.0 ms up)
         Packet loss: 0%

[2025-03-10 20:51:41] Writing data...
[2025-03-10 20:51:41] Saved raw data as crusader-results\test 2025-03-10 20.51.41.crr
[2025-03-10 20:51:41] Saved plot as crusader-results\test 2025-03-10 20.51.41.png

BradenHeat · Mar 10, 2025

Someone from Gl Inet

: https://www.gl-inet.com/blog/how-to-reduce-bufferbloat-with-sqm-on-glinet-routers/

might have missed it on your updated posts, but here i am

xiaofan · Mar 10, 2025

BradenHeat said:
Someone from Gl Inet

: https://www.gl-inet.com/blog/how-to-reduce-bufferbloat-with-sqm-on-glinet-routers/

might have missed it on your updated posts, but here i am

Sorry but I am not so sure why you post this link which is not really relevant to my tests. My main focus is wireless latency test, especially for WiFi 7 router/client with and without using MLO.

There are such tests done before here. Most of them are using WiFi 5/6 routers/clients. I am trying to do more tests on WiFi 7 routers/clients.
https://github.com/Zoxc/crusader/discussions

You are welcome to carry out and share your test results here or in the above github page, using Crusader tool.

1) waveform.com bufferbloat test has proved itself to be basically useless for Singtel users. Crusader is much better test tool to use.
2) More importantly, SQM is not the right tool to help wireless latency. You can say it is the prerequsite to use SQM or other ways to get wired latency issue fixed first. So indeed it may be useful. But then AQL/airtime fairness is the way to help out on wireless client latency, as per the experts. I had a long discussions with two OpenWRT experts on Bufferbloat here.
https://github.com/lynxthecat/cake-autorate/issues/274

xiaofan · Jul 19, 2025

Waveform.com speed has indeed improved the speed for Singtel users. Somehow the wireless speed is higher than OOkla SpeedTest results.

Singtel ONR bridged --> OpenWRT main router --> Switch --> Asus TUF-BE6500 --> wireless --> Acer Swift Go 14 2024 model (Intel BE200 based WiFi 7 adapter)

Main MLO network (2.4GHz + 5GHz MLO). 3m distance.

Bash:

PS C:\work\networking\ookla-speedtest-1.2.0-win64> .\speedtest.exe -s 13623

   Speedtest by Ookla

      Server: Singtel - Singapore (id: 13623)
         ISP: Singtel Fibre
Idle Latency:     3.73 ms   (jitter: 0.50ms, low: 3.42ms, high: 4.50ms)
    Download:  1712.32 Mbps (data used: 2.5 GB)
                 14.54 ms   (jitter: 4.53ms, low: 4.97ms, high: 27.55ms)
      Upload:  1121.67 Mbps (data used: 1.7 GB)
                  9.61 ms   (jitter: 3.99ms, low: 3.55ms, high: 22.91ms)
 Packet Loss:     0.0%
  Result URL: https://www.speedtest.net/result/c/ce48fe4d-0920-46dd-b94a-5bc4c078574b

Ping value from Singtel is also lower than last time.

Bash:

PS C:\work\networking\ookla-speedtest-1.2.0-win64> ping waveform.com

Pinging waveform.com [104.26.1.157] with 32 bytes of data:
Reply from 104.26.1.157: bytes=32 time=5ms TTL=53
Reply from 104.26.1.157: bytes=32 time=5ms TTL=53
Reply from 104.26.1.157: bytes=32 time=5ms TTL=53
Reply from 104.26.1.157: bytes=32 time=6ms TTL=53

Ping statistics for 104.26.1.157:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 5ms, Maximum = 6ms, Average = 5ms

Waveform.com test result:
https://www.waveform.com/tools/bufferbloat?test-id=906ffe7f-f095-41b2-a7a9-e10da0e66fe4

xiaofan · Jul 19, 2025

IoT network (5GHz band)

Bash:

PS C:\work\networking\ookla-speedtest-1.2.0-win64> .\speedtest.exe -s 13623

   Speedtest by Ookla

      Server: Singtel - Singapore (id: 13623)
         ISP: Singtel Fibre
Idle Latency:     2.94 ms   (jitter: 0.25ms, low: 2.73ms, high: 3.27ms)
    Download:  1470.93 Mbps (data used: 2.0 GB)
                 10.21 ms   (jitter: 3.75ms, low: 3.22ms, high: 21.21ms)
      Upload:  1217.81 Mbps (data used: 795.7 MB)
                  9.34 ms   (jitter: 3.15ms, low: 3.05ms, high: 17.07ms)
 Packet Loss:     0.0%
  Result URL: https://www.speedtest.net/result/c/d053e928-44de-4c81-aed4-8b468e7c6060

https://www.waveform.com/tools/bufferbloat?test-id=c652a58d-8627-4ee8-a0c9-e18c4cf0200f

xiaofan · Aug 3, 2025

xiaofan said:
Waveform.com speed has indeed improved the speed for Singtel users. Somehow the wireless speed is higher than OOkla SpeedTest results.

Singtel ONR bridged --> OpenWRT main router --> Switch --> Asus TUF-BE6500 --> wireless --> Acer Swift Go 14 2024 model (Intel BE200 based WiFi 7 adapter)

Main MLO network (2.4GHz + 5GHz MLO). 3m distance.

Waveform.com test result: B
https://www.waveform.com/tools/bufferbloat?test-id=906ffe7f-f095-41b2-a7a9-e10da0e66fe4

Waveform.com Test result this morning for the same setup: A
https://www.waveform.com/tools/bufferbloat?test-id=ea4a803a-e153-4797-9574-2b0f906981a1

lobukong · Aug 3, 2025

xiaofan · Aug 3, 2025

Really sad to hear about the passing of Dave Täht.
https://en.wikipedia.org/wiki/Dave_Täht

xiaofan · Aug 3, 2025

I was wondering why Crusader server at singapore.starlink.taht.net is down. Now I understand. I used this server for some Bufferbloat related testing using Crusader.

Reference:
https://github.com/Zoxc/crusader/issues/9
https://github.com/Zoxc/crusader/discussions/89

xiaofan · Jan 23, 2026

xiaofan said:
Waveform.com Test result this morning for the same setup: A
https://www.waveform.com/tools/bufferbloat?test-id=ea4a803a-e153-4797-9574-2b0f906981a1

Same setup as before.

Singtel ONR bridged --> OpenWRT main router --> Switch --> Asus TUF-BE6500 in AP mode --> wireless --> Acer Swift Go 14 2024 model (Intel BE200 based WiFi 7 adapter)

Main MLO network (2.4GHz + 5GHz MLO), WiFi 7, 4m distance.
https://www.waveform.com/tools/bufferbloat?test-id=363d3e1d-438c-4da9-9c78-e0ba6987c12f

IoT network, 5GHz, WiFi 6

xiaofan · Jan 23, 2026

Singtel ONR bridged --> OpenWRT main router --> Switch --> Asus RT-AX86U in AP mode --> wireless --> Acer Swift Go 14 2024 model (Intel BE200 based WiFi 7 adapter)

Main 5GHz network, WiFi 6. 3m distance.

2.4GHz SSID, WiFi 6, pretty bad upload latency

FYI/A: Bufferbloat 101

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member