let us know something as soon u can broth
Bufferbloat on FTTH
Re: Bufferbloat on FTTH
The 1Gib gives me A or even B on Waveform bufferbloat. If happens even without any external device connected. The only way to get A+ with 0 additional latency under load is with QOS and a moderate cpu usage in the router.triplese wrote: ↑05 Nov 2022, 19:24On 1gbit/s speeds you dont need QoS at all.andrelip wrote: ↑05 Nov 2022, 15:06It's similar to my connection (Vodafone 1Gb) receiving A or B.
I had a second router, Ac68u, that runs the custom firmware (Merlin) that supports fq_codel.
I noticed that for 1Gb, the router could not sustain the QoS and the CPU reached 100%. I had to reduce the bandwidth to 25Mb, and now I can have A+ with 0ms.
And any consumer router cant do this at 2022, even 4-8 core ~2ghz dont enough for that packet rate. You can QoS about 250-300 mbit if your router can process packets in multicore.
Here I disable the Wifi of the ISP’s router and use the second router. Then I can easily switch QOS and VPN on/off.
A curious thing in my router is that the thing that consumed most of the CPU was the algorithm to detect the best Wifi channel. Changing from auto to a hardcoded value cut about 30% of the load.
Re: Bufferbloat on FTTH
I do not mind sharing information, but there are so many contradictions and different opinions on the topic.
At any time of the day or night, no matter what I do, the jitter does not change.
Last edited by 1000WATT on 27 Nov 2022, 05:06, edited 1 time in total.
I often do not clearly state my thoughts. google translate is far from perfect. And in addition to the translator, I myself am mistaken. Do not take me seriously.
Re: Bufferbloat on FTTH
Initially, there should be no packet loss between the computer and your router.espresso wrote:
Why isn't anyone talking about this?
If there are losses on intermediate hosts, but there are no losses on the final ip asdress. This means that on those servers where the losses occur, the firewall is configured to protect against icmp spam. And it doesn't affect the quality.
I often do not clearly state my thoughts. google translate is far from perfect. And in addition to the translator, I myself am mistaken. Do not take me seriously.
Re: Bufferbloat on FTTH
This will be a solution to the problem if the problem is with you, and not with the provider.
SQM is a package for OpenWrt. In different devices, queue management is carried out by different algorithms.
You will have to study the forums before choosing the router that is right for you. A lot of vendors advertise this feature, but some algorithms do better than others using the same hardware.
I often do not clearly state my thoughts. google translate is far from perfect. And in addition to the translator, I myself am mistaken. Do not take me seriously.
Re: Bufferbloat on FTTH
Understood, thanks for the info!
Re: Bufferbloat on FTTH
triplese wrote: On 1gbit/s speeds you dont need QoS at all.
And any consumer router cant do this at 2022, even 4-8 core ~2ghz dont enough for that packet rate. You can QoS about 250-300 mbit if your router can process packets in multicore.
We discussed this a year or two ago.Chief Blur Buster wrote: ↑05 Nov 2022, 20:05You can do efficient gigabit QoS by using two separate routers.
This is indeed a very simple technical solution. But all the same, I want to classify it in the category of "ultimatum".
And yet there are $200 consumer routers that can efficiently manage the queue at 1Gb/s.
Unfortunately, I only have 500MB/s right now. And it will be possible to show efficiency only at this speed, and when I had 1GB / s, only the processor load increased by 85-93%. cake algorithm. The screenshot may appear to be a stitching together of images. Here is a video for 4k monitors.
https://youtu.be/stnDX7zwcog
I often do not clearly state my thoughts. google translate is far from perfect. And in addition to the translator, I myself am mistaken. Do not take me seriously.
Re: Bufferbloat on FTTH
Thanks for your Feedback. But i think the "loss" indications is not my main problem.1000WATT wrote: ↑27 Nov 2022, 04:281.jpgespresso wrote:
Initially, there should be no packet loss between the computer and your router.
Why isn't anyone talking about this?
2.jpg
If there are losses on intermediate hosts, but there are no losses on the final ip asdress. This means that on those servers where the losses occur, the firewall is configured to protect against icmp spam. And it doesn't affect the quality.
I think after years of searching elsewhere i figured out my main problem where i never looked because "i have 1 Gbit fiber".
I can 100% replicate what i see in netgraph when i connect to csgo server.
1. I connect to CSGO server in Frankfurt, my Ping for 1min is 12ms in netgraph
after 1-2 min Ping goes up to mostly 22-23ms
2. I run a trace to the same gameserver via ICMP while my line is Idle
See how the my Average ping is 12ms as well. But you can already see spikes inside my companies network 3. If i now start to add some load to the line the spikes get higher and more consistent and my average PING goes up to 25ms 4. I confirmed in a Swiss forum from another user with the same ISP that he experiences the very same problem
5. I can confirm by connecting directly to the fiber converter without router that the problem is 100% my ISP.
After first hop the latency goes to shit very fast.
Code: Select all
|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| lns21.edge.init7.net - 0 | 3885 | 3885 | 6 | 6 | 7 | 6 |
| r1zrh11.core.init7.net - 0 | 3885 | 3885 | 6 | 7 | 48 | 7 |
| r1zrh7.core.init7.net - 0 | 3885 | 3885 | 6 | 7 | 45 | 7 |
| r1zrh9.core.init7.net - 0 | 3885 | 3885 | 6 | 7 | 89 | 7 |
| r2zrh2.core.init7.net - 1 | 2736 | 2709 | 6 | 7 | 37 | 7 |
| r1zrh2.core.init7.net - 0 | 3885 | 3885 | 7 | 7 | 41 | 7 |
| r1glb1.core.init7.net - 0 | 3885 | 3885 | 7 | 8 | 82 | 7 |
| r1zrh6.core.init7.net - 0 | 3885 | 3885 | 7 | 7 | 39 | 8 |
| r1win12.core.init7.net - 0 | 3885 | 3885 | 7 | 7 | 40 | 7 |
| r2win9.core.init7.net - 0 | 3885 | 3885 | 7 | 7 | 60 | 8 |
| r1win9.core.init7.net - 0 | 3885 | 3885 | 7 | 7 | 20 | 7 |
| r1win7.core.init7.net - 0 | 3884 | 3884 | 7 | 7 | 60 | 8 |
| pspeed01.sys.init7.net - 0 | 3885 | 3885 | 7 | 7 | 8 | 7 |
|________________________________________________|______|______|______|______|______|______|
Re: Bufferbloat on FTTH
That latency going up after some time is weird. Usually it is ping going up during map loading (also depends if map has to be downloaded) and then immediately dropping to low level when map is loaded and stays within 1-2ms variance on good FTTH.
Ryzen 7950X3D / MSI GeForce RTX 4090 Gaming X Trio / ASUS TUF GAMING X670E-PLUS / 2x16GB DDR5@6000 G.Skill Trident Z5 RGB / ASUS ROG SWIFT PG279QM / Logitech G PRO X SUPERLIGHT / SkyPAD Glass 3.0 / Wooting 60HE / DT 700 PRO X || EMI Input lag issue survivor