[SOLVED] DSL reports & bufferbloat

[xyloosepwr]

I tested my internet connection today (1 gbit/s download, 60 mbit/s upload) on dslreports multiple times and i have some confusing results. Bufferbloat results are much better when using VPN (Surfshark). Without VPN my results are: Quality A+, Bufferbloat is B or C sometimes. When connected to VPN, closest server to my location (Frankfurt) bufferbloat is A, Quality is A. Also when connected to VPN my upload speeds are the same like without VPN, only download drops from 1000 mbit/s to 200 mbit/s. I am not really sure should i use VPN for playing online games because of better bufferbloat results on DSLreports?

 

[bill001g]

In many ways that testing should not exist. It confuses people that think red bad green good.

Bufferbloat is a symptom of a problem more than it is a problem itself.

When a connection is overloaded data is placed into buffers. This is done to avoid discarding data. This is actually a good thing for every application except online games and to a lesser extent video conferencing.

The key thing here is data is only placed into buffers when the connection is over utilized. So you never have bufferbloat unless you are using all the bandwidth you purchase. If you are then the solution is to buy more bandwidth. What some people attempt to do instead is use special forms of QoS to try to reduce the buffering of packets when the connection is over used rather than solve the actual problem.

DSL testing intentionally overloads any connection so you will see bufferbloat on every connection in existence. It is a worthless tool unless you are using it correctly. What it is actually used for is to test that you have correctly setup the QoS so that when the connection is overloaded you do not get bufferbloat.

So you first must have a internet connection that is over used. Next you must not be able/willing to upgrade it. And last you need a router that has the fancy QoS that is used to try to solve bufferbloat. Otherwise the bufferbloat tests have no value.

 

[gggplaya]

In many ways that testing should not exist. It confuses people that think red bad green good.

Bufferbloat is a symptom of a problem more than it is a problem itself.

When a connection is overloaded data is placed into buffers. This is done to avoid discarding data. This is actually a good thing for every application except online games and to a lesser extent video conferencing.

The key thing here is data is only placed into buffers when the connection is over utilized. So you never have bufferbloat unless you are using all the bandwidth you purchase. If you are then the solution is to buy more bandwidth. What some people attempt to do instead is use special forms of QoS to try to reduce the buffering of packets when the connection is over used rather than solve the actual problem.

DSL testing intentionally overloads any connection so you will see bufferbloat on every connection in existence. It is a worthless tool unless you are using it correctly. What it is actually used for is to test that you have correctly setup the QoS so that when the connection is overloaded you do not get bufferbloat.

So you first must have a internet connection that is over used. Next you must not be able/willing to upgrade it. And last you need a router that has the fancy QoS that is used to try to solve bufferbloat. Otherwise the bufferbloat tests have no value.

It depends on what you're doing with your internet. Someone viewing a Plex movie outside the house for example. If you look at the chart in this link (View: https://www.reddit.com/r/PleX/comments/hdc6pk/how_can_i_stop_plex_using_high_bandwidth_for_a/
) you can see it spiking to max upload bandwidth when streaming. It's technically not saturating the internet when you take an average speed over several seconds, but these spikes cause buffer bloat because it's saturated momentarily in intervals which can cause lag for someone gaming. So even when you set plex to say 4mbps and you have 12mbps upload. It still does this behavior and spikes to 12mbps in intervals. Even when you set the normal QOS to set plex traffic to the lowest priority and set your gaming computer to the highest priority, you still get lag with simple QOS systems. I know from experience and which is what led me to using FQ_Codel to solve the issue. It really does work, but is a cpu intensive algorithm.

I don't think FQ_Codel will help in the OP's situation. The bufferbloat can be on the VPN server side. Are you actually getting lower latency with a gaming VPN? I tried it once and didn't see any lower latency using a VPN. I guess it all depends on how many connections it takes to get to the gaming server.

 

[SamirD]

In many ways that testing should not exist. It confuses people that think red bad green good.

Bufferbloat is a symptom of a problem more than it is a problem itself.

When a connection is overloaded data is placed into buffers. This is done to avoid discarding data. This is actually a good thing for every application except online games and to a lesser extent video conferencing.

The key thing here is data is only placed into buffers when the connection is over utilized. So you never have bufferbloat unless you are using all the bandwidth you purchase. If you are then the solution is to buy more bandwidth. What some people attempt to do instead is use special forms of QoS to try to reduce the buffering of packets when the connection is over used rather than solve the actual problem.

DSL testing intentionally overloads any connection so you will see bufferbloat on every connection in existence. It is a worthless tool unless you are using it correctly. What it is actually used for is to test that you have correctly setup the QoS so that when the connection is overloaded you do not get bufferbloat.

So you first must have a internet connection that is over used. Next you must not be able/willing to upgrade it. And last you need a router that has the fancy QoS that is used to try to solve bufferbloat. Otherwise the bufferbloat tests have no value.

Or you just have an enterprise router that's waaay overkill so it has enough buffers.

Bill001g has nailed this on the head, and generally I don't care about the bufferbloat result as much as the bandwidth being reported by that test and fast.com.

 

[xyloosepwr]

It depends on what you're doing with your internet. Someone viewing a Plex movie outside the house for example. If you look at the chart in this link (View: https://www.reddit.com/r/PleX/comments/hdc6pk/how_can_i_stop_plex_using_high_bandwidth_for_a/
) you can see it spiking to max upload bandwidth when streaming. It's technically not saturating the internet when you take an average speed over several seconds, but these spikes cause buffer bloat because it's saturated momentarily in intervals which can cause lag for someone gaming. So even when you set plex to say 4mbps and you have 12mbps upload. It still does this behavior and spikes to 12mbps in intervals. Even when you set the normal QOS to set plex traffic to the lowest priority and set your gaming computer to the highest priority, you still get lag with simple QOS systems. I know from experience and which is what led me to using FQ_Codel to solve the issue. It really does work, but is a cpu intensive algorithm.

I don't think FQ_Codel will help in the OP's situation. The bufferbloat can be on the VPN server side. Are you actually getting lower latency with a gaming VPN? I tried it once and didn't see any lower latency using a VPN. I guess it all depends on how many connections it takes to get to the gaming server.

Yes, even latency when using vpn is lower just realized that today. Not a big difference but its lower. Example, in Apex Legends with VPN ping is 11-17ms on Frankfurt servers, without VPN ping is 17-21ms. Bufferbloat results are not always the same. With VPN is always A, without VPN A is only in the morning, in the middle of day and night my results are B or even C sometimes.

 

[xyloosepwr]

Or you just have an enterprise router that's waaay overkill so it has enough buffers.

Bill001g has nailed this on the head, and generally I don't care about the bufferbloat result as much as the bandwidth being reported by that test and fast.com.

https://kabel.vodafone.de/static/media/Technicolor_VodafoneStation_CGA4233DE.pdf

https://www.normann-engineering.com/at/news/news_pdf/0702_DS_Technicolor_CGA4233_EXT_v01.pdf

This is the cable router what i got and only thing that i know that is very limited with settings. I wish to find alternative but they are very rare and very expensive. Connection is mostly used by my PC via LAN cable, TV is also connected on Internet but nobody watch it.

 

[bill001g]

So step back do you have a actual problem or are purely basing all your concern on test results.

Again if you use the DSL testing site it will show bufferbloat on EVERY connection. All this means is you do not have a router with the software to prevent bufferbloat.

 

[SamirD]

Yes, even latency when using vpn is lower just realized that today. Not a big difference but its lower. Example, in Apex Legends with VPN ping is 11-17ms on Frankfurt servers, without VPN ping is 17-21ms. Bufferbloat results are not always the same. With VPN is always A, without VPN A is only in the morning, in the middle of day and night my results are B or even C sometimes.

So one of the ways your latency may be reducing is simply a shorter path via the vpn. Depending on the vpn's peering agreements, they may just have that very unique position of having a better route to the servers than you normally would, and hence better ping.

 

[gggplaya]

But what's your bandwidth when comparing the VPN to your raw connection? It would make sense that you don't get any buffer bloat if the VPN has less bandwidth than your actual raw connection. Your modem's buffers never need to be used.

Most ISP modem/router combos don't have good settings for QOS, they want you to buy more bandwidth anyways. So it would make sense you would get buffer bloat when your connection gets saturated.

With some ISPs, you can set the router/modem into bridge mode so it acts only as a modem. This is the proper way to do it so you won't be double NAT'd. Once you do that, you can attach any router of your choosing which would have better QOS algorithms.

 

[xyloosepwr]

But what's your bandwidth when comparing the VPN to your raw connection? It would make sense that you don't get any buffer bloat if the VPN has less bandwidth than your actual raw connection. Your modem's buffers never need to be used.

Most ISP modem/router combos don't have good settings for QOS, they want you to buy more bandwidth anyways. So it would make sense you would get buffer bloat when your connection gets saturated.

With some ISPs, you can set the router/modem into bridge mode so it acts only as a modem. This is the proper way to do it so you won't be double NAT'd. Once you do that, you can attach any router of your choosing which would have better QOS algorithms.

1000 mbit/s download, 60 mbit/s upload on paper, When testing on speedtest 960 mbit/s download, 58 mbit/s upload. When using VPN upload speeds remain the same, only download speeds drops to 200-300 mbit/s.