Upload vs Download

[sol1109]

I have the AC68U. My service is 15 up and 75 down. I was watching the bandwidth meters when I notice that as two of the PC's in the house began to experience download problems with a game and video a third PC was steadily uploading to YouTube. When his upload was done. the other PC's were able to download data again at full speed. All the PC's are wired. is it normal behavior that upload overrides download or restricts its bandwidth?

 

[Nullity]

Yes, that is very standard behavior which is commonly referred to as bufferbloat.

An early example of using QoS to solve the "uploading ruins download" situation can be found here: http://www.benzedrine.ch/ackpri.html

 

[System Error Message]

The packets that request the download are delayed as they have to be uploaded to request the download. That is why you should set your bandwidth to 90%.

 

[Matthew Patrick]

The packets that request the download are delayed as they have to be uploaded to request the download. That is why you should set your bandwidth to 90%.

On both dl and ul? And are then the bufferfloat gone? Thx

 

[System Error Message]

by setting your bandwidth to 90% you prevent saturation on both upload and download.

 

[Matthew Patrick]

by setting your bandwidth to 90% you prevent saturation on both upload and download.

Yeah. That means i set the limit to 90% for ul and ul. Right? Are the buffer float gonna decrease significantly? Thx

 

[Nullity]

Yeah. That means i set the limit to 90% for ul and ul. Right? Are the buffer float gonna decrease significantly? Thx

Try upload first, then download next. If you have problems, try decreasing the download bitrate, to 60% if needed.

 

[Klueless]

An early example of using QoS to solve the "uploading ruins download" situation can be found here: http://www.benzedrine.ch/ackpri.html

That is a great chart, illustrates problem perfectly. (You're like a human database : -)

 

[Klueless]

The packets that request the download are delayed as they have to be uploaded to request the download.

Great description of the problem.

QoS, bandwidth limiters and/or simple prioritization can all work hand-in-hand towards fixing.

 

[Matthew Patrick]

Okay. So setting download and upload to max 90% my bandwidth. Really makes the ping not high. And please try it by. On the computer . u will try to speedtest so u can see when download. Is the ping high. When upload see if ping increase. So... I use my android terminal to ping to 8.8.8.8 and let it for 1-3 sec. Until i see average ex 30-40 ms. Then. Try the "NOT" limited speedtest. And see for each download or upload. See the ping increase (for me. Download increase from 30ms to 60-70 ms. And upload from 30ms to 200-500ms) so. Then try by limiting it by percentage u want and repeat the test. Until u see stable average ping.

 

[RMerlin]

DSLReport's speed tester is great for this, it tests the ping latency during the upload/download tests. That's how it can evaluate your "buffer bloat index".

 

[Matthew Patrick]

DSLReport's speed tester is great for this, it tests the ping latency during the upload/download tests. That's how it can evaluate your "buffer bloat index".

But when i use it. When i try to limit the inet speed (up/dl) from mac app (so it easier and not time consuming) . the latency is difference. Very difference. So i use the old method that is using my android terminal to ping 8.8.8.8 so i can see the ping result when i limit the bandwidth on mac.

 

[Nullity]

But when i use it. When i try to limit the inet speed (up/dl) from mac app (so it easier and not time consuming) . the latency is difference. Very difference. So i use the old method that is using my android terminal to ping 8.8.8.8 so i can see the ping result when i limit the bandwidth on mac.

I think the dslreports tester uses "in-browser" (javascript?) processing to determine latency, which (for me) does not give accurate results.

The dslreports results add on the lag of browser response time, which is not accurate to true IP packet latency.

 

[Matthew Patrick]

I think the dslreports tester uses "in-browser" (javascript?) processing to determine latency, which (for me) does not give accurate results.

The dslreports results add on the lag of browser response time, which is not accurate to true IP packet latency.

Anyway. I found my bufferfloat. That detected when pinging using my phone and when my computer are on concurrent download or upload. I found out that : (ORIGINALSPEED)-10% = (THEBESTSPEED) . i mean on upload and download. Speed.