FYI/A: Bufferbloat 101

[hwzlite]

I find that www.dslreports.com/speedtest may not able to correctly reflect or identify the bufferfloat as the laptop or browser could be a factor for the spike in ping.

Did a test by performing ping from another machine and running www.dslreports.com/speedtest, when there were ping spikes in the speedtest, there were none to little change in the ping. Even when I ran both the www.dslreports.com/speedtest and Ookla speedtest client at the same time to max out the bandwidth, www.dslreports.com/speedtest showed over 100ms but the ping only increased by 1ms.

You can try enabling the "Hi-Res BufferBloat" option in https://www.dslreports.com/speedtest/preferences .

Ref: [SQM/QOS] Recommended settings for the dslreports speedtest (bufferbloat testing)

 

[yusoffb01]

Why is bufferbloat not issue in my case?

Is it because my isp only 500mbps?
Or huawei Echolife hg8244h modem is good?
Or huawei ax3 router is good?

 

[xiaofan]

Why is bufferbloat not issue in my case?

Is it because my isp only 500mbps?
Or huawei Echolife hg8244h modem is good?
Or huawei ax3 router is good?

Based on my observations, all the above play a part. Your PC also plays a part.

It will be interesting to see the results from SingTel or Whizcomms users which use the HG8244H as an ONR.

 

[hwzlite]

Why is bufferbloat not issue in my case?

Or huawei Echolife hg8244h modem is good?

Huawei Echolife ONT/ONR has build-in QOS feature which its profile might remotely activated by M1 .

I'm on HG8240, which out-of-the-box without any tweaking already give me an "A" rating for bufferbloat

 

[xiaofan]

Interestingly my old Linksys WRT1900AC (April 2014) enables me to get A+ in bufferbloat in different configurations, behind pfsense router (Double NAT) or directly connect to the internet (with public IP). The test is done with Linksys stock FW.

Test three times with the above two configurations, using my Acer Swift 3 (bought in Dec 2017) and Ugreen gigabit Ethernet adapter.
http://www.dslreports.com/speedtest/67585866
http://www.dslreports.com/speedtest/67585879
http://www.dslreports.com/speedtest/67585899

I still get C with my more powerful Asus RT-AX82U router and pfSense Intel Celeron J4105 based router (no QoS applied yet). Upload is pretty bad, probably related to the ZTE ONT.
Asus: http://www.dslreports.com/speedtest/67586285

pfSense: http://www.dslreports.com/speedtest/67586273

 

[xiaofan]

Over wireless connection: the Acer laptop has an upgraded Intel AX200 wifi 6 adapter.

Asus RT-AX82U is not bad today at A.

https://www.dslreports.com/speedtest/67586354

The Singtel Mesh Router (connected to pfSense router as an AP) result is pretty bad at D.

https://www.dslreports.com/speedtest/67586363

Linksys WRT1900AC is still not bad at A.

http://www.dslreports.com/speedtest/67586418

 

[hwzlite]

pfSense: http://www.dslreports.com/speedtest/67586273

Time to enable CoDel Active Queue Management (AQM)

 

[xiaofan]

Time to enable CoDel Active Queue Management (AQM)

Yes that is the plan to enable traffic shaper on pfSense.

But I spent a bit of time on the old WRT1900AC v1 (my main router from April 2014 to August 2020) with openwrt, dd-wrt and the stock Linksys FW, it is strange that the stock FW is actually pretty good in dealing with Bufferbloat. I tried DD-WRT and I need to limit the speed very low (around 200Mbps) to get consistently A grade bufferbloat score which makes no sense to me already. Probably the CPU is not strong enough.

 

[Mach3.2]

http://www.dslreports.com/speedtest/67588635

Ran the test on my new PC, on ethernet same as my last test.

pfsense VM with 2 physical cores allocated > US-24-G1 > R5 5600X w/ an Intel I350-T2 NIC.

 

[xiaofan]

http://www.dslreports.com/speedtest/67588635

Ran the test on my new PC, on ethernet same as my last test.

pfsense VM with 2 physical cores allocated > US-24-G1 > R5 5600X w/ an Intel I350-T2 NIC.

Just wondering if you apply any traffic shaper in this test?

In reality I will try it as a learning experiences but I am not convinced that the traffic shaper is of any real world benefits for 1Gbps connection or even 500Mbps connection.

This article is also interesting.
https://www.speedtest.net/insights/blog/maximized-speed-non-gigabit-internet-connection/#qossettings

 

[Mach3.2]

Just wondering if you apply any traffic shaper in this test?

In reality I will try it as a learning experiences but I am not convinced that the traffic shaper is of any real world benefits for 1Gbps connection or even 500Mbps connection.

This article is also interesting.
https://www.speedtest.net/insights/blog/maximized-speed-non-gigabit-internet-connection/#qossettings

Nope, I didn't apply any traffic shaping as I don't saturate my 500Mbps link regularly.

If the CPU utiliastion on your box is close to maxed out, that could explain the higher load latency.

 

[xiaofan]

Nope, I didn't apply any traffic shaping as I don't saturate my 500Mbps link regularly.

If the CPU utiliastion on your box is close to maxed out, that could explain the higher load latency.

CPU utilization is pretty low for the J4105 pfSense mini PC.

CPU usage :6% to 7% with pfBlockerNG and ntopng in use. I have not checked the usage when running the test but I think it will also be low.

Memory usage:14% of 7969 MiB

SingTel ZTE ONT is the main culprit in this case based on the findings by Miloaisdino. But I think router can help to make the things better.

 

[hwzlite]

http://www.dslreports.com/speedtest/67588635

Ran the test on my new PC, on ethernet same as my last test.

pfsense VM with 2 physical cores allocated > US-24-G1 > R5 5600X w/ an Intel I350-T2 NIC.

Cool.
VM under ESXi / vmxnet3?

 

[Mach3.2]

Cool.
VM under ESXi / vmxnet3?

Yup, that's correct.

The WAN link stays on it's own vSwitch, the LAN link goes into a Distributed vSwitch, load balanced by 2 NICs based on NIC load.

Only time I really use the load balancing is when I'm doing my weekly veeam backup of my VMs, the backup saturates 1 NIC while the routing/internet traffic goes through the other NIC.

 

[miloaisdino]

this looks interesting...

 

[Elijahonli]

Did two rounds of test today

http://www.dslreports.com/speedtest/68424043
http://www.dslreports.com/speedtest/68424106

 

[xiaofan]

Post from Bro BradenHeat
https://forums.hardwarezone.com.sg/...-the-same-condo.6584265/page-3#post-136550158https://www.waveform.com/tools/bufferbloat

 

[xiaofan]

https://www.waveform.com/tools/bufferbloatWe use the following criteria to determine if a particular service will work on your Internet connection. Of course, these criteria are far from perfect, but we think they’re a good general guideline.

• Web Browsing:
  • Download speed > 2 Mbps
  • Upload speed > 100 Kbps
  • Latency < 500 ms
• Audio Calls:
  • Download speed > 100 Kbps
  • Upload speed > 100 Kbps
  • 95th Percentile Latency < 400 ms
• 4K Video Streaming:
  • Download speed > 25 Mbps
• Video Conferencing:
  • Download speed > 10 Mbps
  • Upload speed > 5 Mbps
  • 95th Percentile Latency < 400 ms
• Low Latency Gaming:
  • Download speed > 10 Mbps
  • Upload speed > 3 Mbps
  • 95th Percentile Latency < 40 ms

 

[BradenHeat]

https://www.waveform.com/tools/bufferbloatWe use the following criteria to determine if a particular service will work on your Internet connection. Of course, these criteria are far from perfect, but we think they’re a good general guideline.

• Web Browsing:
  • Download speed > 2 Mbps
  • Upload speed > 100 Kbps
  • Latency < 500 ms
• Audio Calls:
  • Download speed > 100 Kbps
  • Upload speed > 100 Kbps
  • 95th Percentile Latency < 400 ms
• 4K Video Streaming:
  • Download speed > 25 Mbps
• Video Conferencing:
  • Download speed > 10 Mbps
  • Upload speed > 5 Mbps
  • 95th Percentile Latency < 400 ms
• Low Latency Gaming:
  • Download speed > 10 Mbps
  • Upload speed > 3 Mbps
  • 95th Percentile Latency < 40 ms

Protest on the last point : codm on ipad, dont expect miracles as its strictly base on server and other plays using VPN [ aka i dont think theres CodM server in singapore, as im playing with Indo and Malaysia players 90% of the time ]

lol the rest actually, video calls should be dependent on the application their using , e.g Teams is a heavier usage vs something like google rooms / calls

lol but nice guide

Post from Bro BradenHeat
https://forums.hardwarezone.com.sg/...-the-same-condo.6584265/page-3#post-136550158https://www.waveform.com/tools/bufferbloat

aiseh i feel honored sia

 

[firesong]

Not the most exciting results.

Also, tested 10x, but only once this Bufferbloat test loaded. Wonder what gives.