FYI/A: Bufferbloat 101

[xiaofan]

So far based on observation here from those who benchmarked from 10Gbps plan, seems that Waveform.com still has good enough bandwidth to cater those =< 2.5Gbps plan.

IMHO , wireless test will be still meaningful for someone who appreciates consistent real-time buffering/stuttering-free video/chat/stream/vod sesssions ( eg. Teams, Zoom, Google Meet, Whatsapp ..etc any1?), at least can be certain 99% is not localized issue if any of my home dweller nags about it.

We all agree that latency is important. Real bufferbloat situation may be important as well.

The main difference in opinions is whether it is important to look at the score of Waveform.com bufferbloat test results.

Even though has good enough bandwidth to cater those =< 2.5Gbps plan, I tend to think it is only applicable to wired test results.

For the use cases you quote, I do not think waveform.com bufferbloat test is representative at all for most users, no matter wired or wireless, as the video/chat/stream/vod sessions require much less bandwidth than the simulated tests. It may represent users who are carrying out video/chat/stream/vod sessions while downloading a lot of stuff (like doing SpeedTest), which is quite rare.

Edit to add:
But I could be totally wrong here as well. I am more based on my own experiences. I am using the following two home networks. I do not see any differences on either networks for use cases like WFH/HBL (Teams, Google Meet, Zoom), no matter I get an A or a C in the tests.
1) Asus RT-AX86U
2) OpenWRT virtual router + ZTE BE7200 Pro+ AP (previously Asus RT-AX82U AP).

 

[xiaofan]

Similarly the difference in opinions is also related whether QoS/SQM is really important or not for 1Gbps and above.

Typical consumer routers have low CPU performance to carry out 1Gbps QoS/SQM anyway. So I tend to think they do more harm to typical users than helping.

For 10Gbps plan, it is even meaningless to use QoS/SQM as I do not know what kind of CPU can do the job.

But I could be totally wrong here as well. I am more based on my own experiences. I am using the following two home networks. I do not see any differences on the OpenWRT side, with or without SQM. On the Asus side, it would be a disaster to enable Asus Adaptive QoS as the line speed will drop a lot.
1) Asus RT-AX86U
2) OpenWRT virtual router + ZTE BE7200 Pro+ AP (previously Asus RT-AX82U AP).

 

[hwzlite]

It may represent users who are carrying out video/chat/stream/vod sessions while downloading a lot of stuff (like doing SpeedTest), which is quite rare.

That's the point to pro-actively anticipate those "quite rare" events , worries-free with anti-bufferbloat mechanism inplace with a consistent networking experience

 

[xiaofan]

That's the point to pro-actively anticipate those "quite rare" events , worries-free with anti-bufferbloat mechanism inplace with a consistent networking experience

Sorry but I do not agree. Since typical users here will use 1Gbps plan, and anti-bufferbloat mechanism to me does more harm than good for 1Gbps plan for typical consumer routers. It is of course okay if you use an Intel mini PC and running OpenWRT/pfSense with 1Gbps plan. Even in that case, I am skeptical the real usefulness for typical use cases.

Then for 10Gbps plan, it is probably meaningless to do any types of QoS/SQM.

Basically my idea is to get a more powerful router and live with the bufferbloat score of B or A (usually I can get A with wired connection or close range wireless anyway with my 1Gbps plan, without using any QoS/SQM measures), no need to chase A+ using any QoS or SQM measures.

But again, it is just my personal opinion and I could be totally wrong.

Similarly the difference in opinions is also related whether QoS/SQM is really important or not for 1Gbps and above.

Typical consumer routers have low CPU performance to carry out 1Gbps QoS/SQM anyway. So I tend to think they do more harm to typical users than helping.

For 10Gbps plan, it is even meaningless to use QoS/SQM as I do not know what kind of CPU can do the job.

But I could be totally wrong here as well. I am more based on my own experiences. I am using the following two home networks. I do not see any differences on the OpenWRT side, with or without SQM. On the Asus side, it would be a disaster to enable Asus Adaptive QoS as the line speed will drop a lot.
1) Asus RT-AX86U
2) OpenWRT virtual router + ZTE BE7200 Pro+ AP (previously Asus RT-AX82U AP).

 

[xiaofan]

Basically my idea is to get a more powerful router and live with the bufferbloat score of B or A (usually I can get A with wired connection or close range wireless anyway with my 1Gbps plan, without using any QoS/SQM measures), no need to chase A+ using any QoS or SQM measures.

But again, it is just my personal opinion and I could be totally wrong.

Just an example, when the wireless signal is weak, the score will be quite bad.

This is with the same setup as before: Singtel 1Gbps plan, OpenWRT + ZTE BE7200 Pro+ wireless AP, Acer Windows 11 laptop with Intel AX201 WiFi 6 adapter. But now the laptop is two walls away (old HDB flat) at the corner of the kitchen area (close to the Windows), weakest signal area in my 4-room 106sqm HDB flat with thick walls.

No one is using internet there so I do not really care. However, if I need to fix the issue there, what should I do? I do not think using QoS/SQM will help much as the laptop has issues to transmit the signal to the router (upload problem). Rather I should get a mesh node there to help improve the wireless signal if I care for the upload. On the other hand, download is actually quite usable (router has strong signal) -- I just checked and no issues with 4k Youtube.

 

[sglandscape]

Just an example, when the wireless signal is weak, the score will be quite bad.

This is with the same setup as before: Singtel 1Gbps plan, OpenWRT + ZTE BE7200 Pro+ wireless AP, Acer Windows 11 laptop with Intel AX201 WiFi 6 adapter. But now the laptop is two walls away (old HDB flat) at the corner of the kitchen area (close to the Windows), weakest signal area in my 4-room 106sqm HDB flat with thick walls.

No one is using internet there so I do not really care. However, if I need to fix the issue there, what should I do? I do not think using QoS/SQM will help much as the laptop has issues to transmit the signal to the router (upload problem). Rather I should get a mesh node there to help improve the wireless signal if I care for the upload. On the other hand, download is actually quite usable (router has strong signal) -- I just checked and no issues with 4k Youtube.

This seems to be a wireless congestion issue so I'll say don't worry about it. If you're testing using wired and don't get an A then I'll suggest checking your settings.

 

[hwzlite]

I just checked and no issues with 4k Youtube.

Well, not really a fair test to demo its benefit as streaming data is preloaded cached in video player software while watching the video....
Check whether you can disable or reduce buffer cache in video player software for the sake to really show its effect.

 

[xiaofan]

Well, not really a fair test to demo its benefit as streaming data is preloaded cached in video player software while watching the video....
Check whether you can disable or reduce buffer cache in video player software for the sake to really show its effect.

I mean it is a meanlingless test as the wireless signal is low. So no point carrying out any tests.

 

[xiaofan]

This seems to be a wireless congestion issue so I'll say don't worry about it. If you're testing using wired and don't get an A then I'll suggest checking your settings.

I have always been able to get an A with wired test.

In this paticular wireless test case, the signal is weak -- no use with a powerful router as it is the client which can not effectively transmit signals back to the router -- so you can see upload is bad.

Will SQM/QoS help? I believe NOT.

 

[Mach3.2]

no use with a powerful router as it is the client which can not effectively transmit signals back to the router

you found one of the reasons for deploying multiple AP setups.

 

[cyberet]

my bufferbloat score is subjected to my neighbours
when they go to work, my score is A.
when they are back home, my score is C.

wireless in high rise enviroment is

 

[hwzlite]

I will leave the truly scientific test from the pro, Google Engineers on "Airtime Based Queue Limit for FQ-CoDel in Wireless Interfaces" test on Google WIFI (802.11ac) , especially interesting observation on its "Test Result > Rate vs Range (RvR) test" topic :

""At higher attenuation, the improvement is even more significant; there is more than 10x latency reduction at the 40 dB attenuation point.""

Ofcos don't expect its mitigation is a silver bullet for everything hor

 

[hwzlite]

my bufferbloat score is subjected to my neighbours
when they go to work, my score is A.
when they are back home, my score is C.

wireless in high rise enviroment is

Take a chance @ DFS channels ?

 

[BradenHeat]

Eh must be y'all test test test, until Red Sea buay tahan cut cut cut, now FB all pai liao la

 

[cyberet]

Take a chance @ DFS channels ?

stay too near airport, keep getting kick off DFS

 

[xiaofan]

Update on my WIRELESS adventure into using fq_codel/ATF/AQL optimized MediaTek mt76 mac80211 driver on Linksys EA8100(OpenWRT) , configured as Dumb AP, and tweaked with "aql_txq_limit = 1500" setting :

Here are the wireless benchmarks (M1 500Mbps, DFS channel 64 @80Mhz, Wall = old HDB type):
Line-Of-Sight, ~6m:
https://www.waveform.com/tools/bufferbloat?test-id=ef0c1814-5699-463b-957c-0b057da1e06b (A+)
https://www.waveform.com/tools/bufferbloat?test-id=9669484d-82ab-4561-acf3-e040a15ea052 (A+)
https://www.waveform.com/tools/bufferbloat?test-id=9e60560f-cc01-4ede-b17d-63d7b3fc6065 (A+)

Block by 1x Wall , ~8m:
https://www.waveform.com/tools/bufferbloat?test-id=653bcd05-58c3-4f4d-9d15-0c97a004f7fd (A)
https://www.waveform.com/tools/bufferbloat?test-id=9b69cf66-f6d0-4305-9b38-f5fbf7b98dc7 (A)
https://www.waveform.com/tools/bufferbloat?test-id=336b7e85-4913-4367-b8a3-4c3910fee3d5 (A)

Block by 2x Walls, ~9m:
https://www.waveform.com/tools/bufferbloat?test-id=6ea67cfd-1681-4f21-974a-246094a7181a (B)
https://www.waveform.com/tools/bufferbloat?test-id=54d11a51-bd0c-4a49-9016-2a3341b0f569 (B)
https://www.waveform.com/tools/bufferbloat?test-id=5a0d52fb-0192-41cb-8931-fbe87f422938 (C)


Ref: How OpenWrt Vanquishes Bufferbloat

If we look at the results, once with wireless signal drops, the score gets down from A+ to A to B to C.

From the speed drop, this looks like the wall is not blocking WiFi so badly as my older 1998 HDB.

In my case, it already drops to F with one thick wall. In my paticular case, I do not worry too much about the weak signal area (kitchen/common toilet/service yard) since I still have pretty good speed in the area. Our two Windows laptops, one Chromebook and Mac Mini M1 are located in the living room and Common Room 1, all these devices will have pretty fast wireless speed for both download and upload (>=500Mbps).

Tests done this morning.
Singtel 1Gbps plan, OpenWRT + ZTE BE7200 Pro+ wireless AP, Acer Windows 11 laptop with Intel AX201 WiFi 6 adapter

Living room -- 4m away (A)
https://www.waveform.com/tools/bufferbloat?test-id=ca28094a-9e60-4ebc-991f-b07f28d57353 (A)
https://www.waveform.com/tools/bufferbloat?test-id=a60ad081-f0f4-4a85-a6ee-13f9005693ff (A)

Common Room 1 -- 6m away, door closed, one wall (not thick wall) in between (A)
https://www.waveform.com/tools/bufferbloat?test-id=ce6476bb-6aa6-416c-b155-37b25ceb2461 (A)

Common toilet -- 7m away, one thick wall (F)
https://www.waveform.com/tools/bufferbloat?test-id=aa70abc4-0c9b-44ca-8045-dc10a764a784 (F)

Yard area -- 9m away, one more wall after the common toilet (not thick wall) (F)
https://www.waveform.com/tools/bufferbloat?test-id=c5fd6391-968c-4306-b016-8778dad4a78b (F)

 

[xiaofan]

I will test in the evening as well where the interferences may be more prominent.

In this particular case, since I am using OpenWRT, I can carry out experiments using various SQM measures.

For my other Asus RT-AX86U, I do not really have the luxury in reality since the CPU is weak to carry out meaningful QoS for 1Ghz Fibre Internet. I know people fond of Merlin FW will say otherwise...

 

[xiaofan]

I will test in the evening as well where the interferences may be more prominent.

1) No much differences in the evening (about 10:30pm). This time I was testing with the Asus RT-AX86U router.
Singtel 1Gbps plan, Asus RT-AX86U wireless router, Acer Windows 11 laptop with Intel AX201 WiFi 6 adapter

Living room -- 4m away (A)
https://www.waveform.com/tools/bufferbloat?test-id=adffe778-ddee-406c-8040-3fbece7544d9

Common toilet -- 7m away, one thick wall (F)
https://www.waveform.com/tools/bufferbloat?test-id=37ce8292-33aa-485a-b24b-5d63fa48001a

Yard area -- 9m away, one more wall after the common toilet (not thick wall) (F)
https://www.waveform.com/tools/bufferbloat?test-id=1ff9c384-ff0b-4bb3-8259-2b8b71fb8e17

2) Edit to add: just tested OpenWRT + ZTE BE7200 Pro+ as well (around 10:40pm). No much difference as well. Looking at the detailed results, maybe slightly better than Asus RT-AX86U

Living room -- 4m away (A)
https://www.waveform.com/tools/bufferbloat?test-id=50b4f7c6-755a-4cbf-9631-75396c599170

Common toilet -- 7m away, one thick wall (F)
https://www.waveform.com/tools/bufferbloat?test-id=78987526-71fe-4f81-83d5-75e75d7a015a

Yard area -- 9m away, one more wall after the common toilet (not thick wall) (F)
https://www.waveform.com/tools/bufferbloat?test-id=b9bc4452-217c-4b67-94b6-122b4624da86

 

[hwzlite]

....From the speed drop, this looks like the wall is not blocking WiFi so badly as my older 1998 HDB.

Mine is the "squarish" 4-Rm Model A type built on 89.

Getting a little scientific with visualization from Heat Map generated by WiFi Analyzer and Surveyor :

(-25 dBm = DumbAP location)

By Agaration, can derive that anything <55dBM signal strength get "A/A+" rating.

Worse "F" hits are from the Kitchen's toilet, largely bogged down by upload latencies:
https://www.waveform.com/tools/bufferbloat?test-id=4eea6f46-8bed-4717-9e1a-d4ad89e54973
https://www.waveform.com/tools/bufferbloat?test-id=019de526-26bb-4f53-a236-ea5cfc1ca869
https://www.waveform.com/tools/bufferbloat?test-id=872038ef-9cce-4884-80aa-586e80ffa56a

 

[xiaofan]

Mine is the "squarish" 4-Rm Model A type built on 89.

Getting a little scientific with visualization from Heat Map generated by WiFi Analyzer and Surveyor :

(-25 dBm = DumbAP location)

By Agaration, can derive that anything <55dBM signal strength get "A/A+" rating.

Worse "F" hits are from the Kitchen's toilet, largely bogged down by upload latencies:
https://www.waveform.com/tools/bufferbloat?test-id=4eea6f46-8bed-4717-9e1a-d4ad89e54973
https://www.waveform.com/tools/bufferbloat?test-id=019de526-26bb-4f53-a236-ea5cfc1ca869
https://www.waveform.com/tools/bufferbloat?test-id=872038ef-9cce-4884-80aa-586e80ffa56a

Nice signal map. Now your results are matching my expectations.

Your results show a direct relation of waveform.com bufferbloat test results with the signal strength. And indeed upload latency is the main issue as the client devices like mobile phones or laptops will usually have lower transmission power than the router -- lower speed and higher latency for the upload compared to the download side when signal is not good.

Luckily download speed and latency will be more important than upload for majorties of the applications.

Still in your posted results, both download speed and upload speed are low, this shows a weak router as well. And I know that is because you are using an old Linksys EA8100 v1 router.

My 106 sqm HDB flat was built in 1998. The "F" results posted previously were from Point D and E. Point D is probably the area with weakest signal. My download speed is still quite decent, with either Asus RT-AX86U router or ZTE BE7200 Pro+, both are much more powerful than the poor Linksys EA8100v1.

Point D waveform.com bufferbloat "F" results posted before, using Asus RT-AX86U and OpenWRT+ZTE BE7200 Pro+ AP respectively. You can see the download and upload speed are still quite good, but the upload latency is bad under load (weak signal and the client has low Tx power to send signal back to the router).

https://www.waveform.com/tools/bufferbloat?test-id=b9bc4452-217c-4b67-94b6-122b4624da86
https://www.waveform.com/tools/bufferbloat?test-id=c5fd6391-968c-4306-b016-8778dad4a78b