Why is my Anti-Bufferbloat not working?

[vShikko]

 

Well I have had the XR500 for well over a month now, and I have waited till now to mention it. As I wanted to test over a period of time, and see what result's I come up with.

 

I have noticed when using the dslreports speedtest, my result's are better without "Anti-Bufferbloat", then with it enabled. Which honestly is a bit backwards, and should be the opposite. I have run many of tests the past month, or so, and everytime my bufferbloat results are better when I leave my speeds uncapped. I plan to post the results of two recent tests in this post. One without it, and the other with my speed's capped to 25% of my overall total bandwidth. As you will see bufferbloat isn't exactly improving "bufferbloat" when enabled.

 

I actually recently been paying close attention to the test being run, and with Anti-bufferbloat enabled. During the upload side of the testing, the latency average increases, spikes a lot more often, and higher. Which when you think about it.. With me having my limit's set to 25%, I can only use a max of 7.5-7.75mbps of upload bandwidth. I actually get 30-32mbps by my provider. So overall, latency, and ping spikes should stay very low, and consistent. Currently that isn't the case, and I feel whatever the issue is, it's router side.

 

Because testing without it, latency, and spikes, are much less. This is while maxing out my upload at 30-32mbps, which would be the time for my latency to be at its highest, and spike the most. However that isn't the case. So at this point in time, I feel using "Anti-Bufferbloat" is a waste, and isn't doing exactly what it's designed to do.

 

 

Why are you throttling your bandwidth to 25%?

You're making your router queue up more packets than it has to & creating a bottleneck.

[Find_the_Door]

Fraser,

 

I will get back to you as soon as I can. But until then, I plan to dig into this more from my end as well. This has me very curious as to why it's happening. dslreports speedtest bufferbloat part of the test depends on result's it collects while stressing your bandwidth. Oddly for the upload side, as you have seen... The result's are worst while Anti-BB is enabled. Which doesn't make a whole lot of sense, as I'm only using 25% of my upload with it enabled, vs 100% without it.

 

I'm very curious why its shows latency average higher, and with higher spiking. With Anti-BB enabled, is the router cpu strictly doing all the processing? I know often times certain features being enabled, disables other stuff, which normally keeps the cpu load down.

 

Did you ever find a solution here?  My results are very similar to yours.

[Netduma Fraser]

Did you ever find a solution here?  My results are very similar to yours.

 

I'll hopefully be doing a 1 on 1 this week.

 

Have you done the same tests? Are ALL devices connected to the XR500? Otherwise QoS won't work correctly.

[Find_the_Door]

I'll hopefully be doing a 1 on 1 this week.

 

Have you done the same tests? Are ALL devices connected to the XR500? Otherwise QoS won't work correctly.

Yes they're all connected. I have to significantly neuter my connection to get decent results. I have Gigabit with 42 up.

 

Final numbers are around 250 down and 29 up to get it working right somewhat. I notice under load my ping time is actually better? Is there a reason behind this? Is there a way I can enforce that better ping time consistently?

[Netduma Fraser]

That's probably to do with your connection speed. So under load with Anti-Bufferbloat active you get a better ping? If so just keep it like that. Otherwise that doesn't make any sense why it would be lower

[Find_the_Door]

That's probably to do with your connection speed. So under load with Anti-Bufferbloat active you get a better ping? If so just keep it like that. Otherwise that doesn't make any sense why it would be lower

Yeah I'm definitely gonna keep it that way for sure. For whatever reason it does that when I watch the in-game network monitor on Rainbow Six.

[Netduma Fraser]

Yeah I'm definitely gonna keep it that way for sure. For whatever reason it does that when I watch the in-game network monitor on Rainbow Six.

 

Well that is great news! If you get your best results at that percentage thats great, I realise it's a big drop in speed but everyones connection is different. Glad you found what works for you!

[Find_the_Door]

Well that is great news! If you get your best results at that percentage thats great, I realise it's a big drop in speed but everyones connection is different. Glad you found what works for you!

It works for now for sure - I'll continue to monitor it going forward. I would like an option to disable the deep packet filtering if possible. If you had to provide a specialized forum only firmware to do so that's fine, or if you needed to hid the option away in the GUI whatever it takes.

 

I'd say it'd save you guys a lot of time and energy as far as devoting discussion's to this issue with regards to speed test results. More importantly it'd make optimization that much easier.

 

Just curious what queing discipline does Netduma OS use? How about the R1? If it's FQ_Codel is there an option to control packet overhead?

 

 

Thank you

[Netduma Fraser]

The only option to disable that at the moment is when you disable QoS completely. We may in future offer an option to disable DPI but not for awhile. 

 

We use our own as far as I'm aware, no options for that either. We have to be careful implementing advanced features like that.

[Find_the_Door]

The only option to disable that at the moment is when you disable QoS completely. We may in future offer an option to disable DPI but not for awhile.

 

We use our own as far as I'm aware, no options for that either. We have to be careful implementing advanced features like that.

Roger that - any specific option for overhead? I seen something called "goodput" does that have any particular effect on the experience?

 

Also when should we expect our next update?

 

Thank you

[Netduma Jack]

Roger that - any specific option for overhead? I seen something called "goodput" does that have any particular effect on the experience?

 

Also when should we expect our next update?

 

Thank you

 

Goodput should only be disabled in very specific circumstances and I don't believe it would cause any performance changes either way. The next update should be fairly soon but we can't give a date yet - stay tuned on this forum and also on our social media channels for news regarding that

[Netduma Fraser]

Roger that - any specific option for overhead? I seen something called "goodput" does that have any particular effect on the experience?

 

Also when should we expect our next update?

 

Thank you

 

Goodput accounts for packet headers so that speedtest gives a more accurate result

[xlr8r]

managed to get my first A+ Bufferbloat result on DSLREPORTS

 

  

 

My Normal DL is 64Mbps , UL 12Mbps (while QoS is turned Off)

 

For this test, I had QoS turned on and I set my sliders to 75 DL and 73 UL (after testing many, many  times on DSLReports)

 

now getting an A+ Bufferfloat

[Find_the_Door]

managed to get my first A+ Bufferbloat result on DSLREPORTS

 

 

My Normal DL is 64Mbps , UL 12Mbps (while QoS is turned Off)

 

I set my sliders to 75 DL and 73 UL (after testing many, many times on DSLReports)

 

now getting an A+ Bufferfloat

Did this suddenly happen or did you further tinker with settings to arrive at this result?

[xlr8r]

Did this suddenly happen or did you further tinker with settings to arrive at this result?

 

Spent all day yesterday doing this..... its a tad inconsistent , but at least I managed to achieve an A+  (maybe it was just a fluke...?) 

Firstly, (with QoS off) I obtained my accurate DL and UL speeds first by setting my Anti-bufferfloat internet speeds to way above what i should be getting. Then ran a speedtest to get latest speeds.

Entered those back into the antibufferfloat internet speed settings on XR500, and saved. Turned QoS back on for testing.

I then tinkered ALOT with dsl reports (morning / noon/ night) to obtain what the best settings where (for me).

I tested all the way down to 40% to 100% on both sliders , and throughout the day, tested many times on various slider values.

 

Ultimately, the best values for me, were 75 DL and 73 UL on the sliders , with QoS enabled (always - just for testing purposes)

Credits to SIMJC74 who has a nice video of how to do the Bufferbloat testing..... https://www.youtube.com/watch?v=p41FUBTNRHM

 

Speedtest 1 here (http://www.speedtest.net) (internet speed)

Speedtest 2 here  (https://sourceforge.net/speedtest/)  (inlcudes Latency/Ping, Jitter, Download Speed, Upload Speed, Buffer Bloat, and Packet Loss)

DSLReports here (http://www.dslreports.com/speedtest) (includes bufferfloat tests)

[lilstone87]

Well I have did more testing, and digging on this, the last couple days. I noticed a couple things, and I'm gonna share my findings below.

 

 

1. Oddly at times I noticed my result's were not the best while using google chrome. I used firefox, and things seemed more steady. I have been able to get some good result's with chrome as well though. So I wouldn't say not use it. I tried using windows 10 built in IE browser, however I tried a couple test's, all failed when trying to do the upload side of the test. So I say stay away from IE for testing.

 

2. When running this test, I strongly suggest cutting all bandwidth usage on the device you're testing on. So if you're on your pc, which you shouldn't be testing over a wifi device anyways. So testing while on a PC/Laptop. Close any apps/browser tabs that are using active bandwidth till testing is over. Also I highly recommend when testing on dslreports.com, try to only have it open within the browser you're using. I feel that will minimize any other tabs within the browser somehow causing the test an issue.

 

3. Do yourselves a favor, and create an account on dslreports.com, will only take a short time to do so. Reason being, you can configure your "Preferences" settings for the speedtest, once saved, they will stay linked to your account. I feel configuring the preference settings is VERY important with this speedtest. I will explain more below.

 

4. Reason to setup preference settings. First you can select the servers you want to use, for the test. At the top you have a option to let it pick some servers for you based off of either Latency, or Bandwidth. I suggest using the "Latency" option. After that's done running, it will have the servers it picked with a check on them. After that, you can look at the one's it picked, and unselect some of them if you want. I did a couple for myself, pick the ones closest to me. After this, make sure to hit the save button at the bottom.

 

However I'm not done with the preference settings yet... After getting a small number of servers picked based off latency by it's own testing, and also checking over them yourself. I wouldn't have no more then 8-10 servers selected.

 

Now scroll down a little, and you will see "No. download streams" with a box beside it. You can pick between 1-32. I honestly think for most of us "4" will be enough. That will be the amount of servers it will use while testing. Do the same for the upload side, which is right below it.

 

For awhile I was using 16+ for each test I done, and I now don't recommend using that high of a number of servers for each test. Reason being, your overall result's will depend on all the servers used during that test. For me, I'm on the east coast of the US, and when I'm doing one of these test using 16+ servers. I'm likely using several servers on the west coast of the US. With a mixature of servers all over, there's so many things that can affect your routing from one minute to the next.

 

So my final opinion on this, keep the number of servers used for the test low, and use servers close to you. Because this to me will give you more accurate result's, when it comes to the bufferbloat part of this test. Finally make sure after any preference setting changes, you make sure to click the "save" button at the bottom.

 

 

 

5. After doing the stuff above, I have been seeing more steady result's with the dslreports.com speedtest, in terms of bufferbloat, and my overall results. I don't feel the current default settings used for their speedtest, is optimized very well for testing, and getting the most accurate results. Ultimately I think by default they're running the test using to many servers, which can be from far away. With that, comes a lot of possible routing issue's. Some of them you have no control over, and isn't an issue with your connection. However whatever each server is reporting back, ultimately is included at the end of the test, and will factor into your final result.

[Alex49H]

Well I have did more testing, and digging on this, the last couple days. I noticed a couple things, and I'm gonna share my findings below.

 

 

1. Oddly at times I noticed my result's were not the best while using google chrome. I used firefox, and things seemed more steady. I have been able to get some good result's with chrome as well though. So I wouldn't say not use it. I tried using windows 10 built in IE browser, however I tried a couple test's, all failed when trying to do the upload side of the test. So I say stay away from IE for testing.

 

2. When running this test, I strongly suggest cutting all bandwidth usage on the device you're testing on. So if you're on your pc, which you shouldn't be testing over a wifi device anyways. So testing while on a PC/Laptop. Close any apps/browser tabs that are using active bandwidth till testing is over. Also I highly recommend when testing on dslreports.com, try to only have it open within the browser you're using. I feel that will minimize any other tabs within the browser somehow causing the test an issue.

 

3. Do yourselves a favor, and create an account on dslreports.com, will only take a short time to do so. Reason being, you can configure your "Preferences" settings for the speedtest, once saved, they will stay linked to your account. I feel configuring the preference settings is VERY important with this speedtest. I will explain more below.

 

4. Reason to setup preference settings. First you can select the servers you want to use, for the test. At the top you have a option to let it pick some servers for you based off of either Latency, or Bandwidth. I suggest using the "Latency" option. After that's done running, it will have the servers it picked with a check on them. After that, you can look at the one's it picked, and unselect some of them if you want. I did a couple for myself, pick the ones closest to me. After this, make sure to hit the save button at the bottom.

 

However I'm not done with the preference settings yet... After getting a small number of servers picked based off latency by it's own testing, and also checking over them yourself. I wouldn't have no more then 8-10 servers selected.

 

Now scroll down a little, and you will see "No. download streams" with a box beside it. You can pick between 1-32. I honestly think for most of us "4" will be enough. That will be the amount of servers it will use while testing. Do the same for the upload side, which is right below it.

 

For awhile I was using 16+ for each test I done, and I now don't recommend using that high of a number of servers for each test. Reason being, your overall result's will depend on all the servers used during that test. For me, I'm on the east coast of the US, and when I'm doing one of these test using 16+ servers. I'm likely using several servers on the west coast of the US. With a mixature of servers all over, there's so many things that can affect your routing from one minute to the next.

 

So my final opinion on this, keep the number of servers used for the test low, and use servers close to you. Because this to me will give you more accurate result's, when it comes to the bufferbloat part of this test. Finally make sure after any preference setting changes, you make sure to click the "save" button at the bottom.

 

 

 

5. After doing the stuff above, I have been seeing more steady result's with the dslreports.com speedtest, in terms of bufferbloat, and my overall results. I don't feel the current default settings used for their speedtest, is optimized very well for testing, and getting the most accurate results. Ultimately I think by default they're running the test using to many servers, which can be from far away. With that, comes a lot of possible routing issue's. Some of them you have no control over, and isn't an issue with your connection. However whatever each server is reporting back, ultimately is included at the end of the test, and will factor into your final result.

 

That's good info Scott!!! Thanks!

[lilstone87]

Here's a test I just did, to show I'm not just posting what I feel might be right. I forgot to attach it to my post from a few minutes ago.

 

 

EDIT:

Here's another test I just did from a minute ago, where I severely limited my speeds, I still got "A" for bufferbloat. Before best I seen was "B" for bufferbloat. Disregard the "F" for speed, as that's based off my average which would normally be 330/30. But I limited my speed for this test, and a little bit for the test above as well.

 

[Netduma Fraser]

Nicely done Scott, glad to see you found what may be the cause to bufferbloat readings being lower than they should be! 

[Alex49H]

I have made the adjustments you suggested Scott and now I get an A+ while testing.

[lilstone87]

Good to hear Alex. I spent a couple days messing around with things on my end, and finally on the test's end with settings. Plus i overlooked many test result's, before i decided to post these findings.

 

I'm not one who will post about something, unless i feel 100% confident it's true. So like i said i took my time on this, and honestly probably burnt threw 10-15gb of data over a couple day period, just running this test repeatedly.

 

Ultimately i'm happy you were able to take the info i provided, put it to use, and get the same kind of results i got. Which are what people want to see, if their connection is running optimally.

 

Finally i hope others who are having trouble getting results they want with this test, see my earlier post, and are able to put my suggestions to good use. As i think the test can be good to use, it just isn't optimized to provide accurate results, when using its default settings.

[Soloflyerz]

Well I have did more testing, and digging on this, the last couple days. I noticed a couple things, and I'm gonna share my findings below.

 

 

1. Oddly at times I noticed my result's were not the best while using google chrome. I used firefox, and things seemed more steady. I have been able to get some good result's with chrome as well though. So I wouldn't say not use it. I tried using windows 10 built in IE browser, however I tried a couple test's, all failed when trying to do the upload side of the test. So I say stay away from IE for testing.

 

2. When running this test, I strongly suggest cutting all bandwidth usage on the device you're testing on. So if you're on your pc, which you shouldn't be testing over a wifi device anyways. So testing while on a PC/Laptop. Close any apps/browser tabs that are using active bandwidth till testing is over. Also I highly recommend when testing on dslreports.com, try to only have it open within the browser you're using. I feel that will minimize any other tabs within the browser somehow causing the test an issue.

 

3. Do yourselves a favor, and create an account on dslreports.com, will only take a short time to do so. Reason being, you can configure your "Preferences" settings for the speedtest, once saved, they will stay linked to your account. I feel configuring the preference settings is VERY important with this speedtest. I will explain more below.

 

4. Reason to setup preference settings. First you can select the servers you want to use, for the test. At the top you have a option to let it pick some servers for you based off of either Latency, or Bandwidth. I suggest using the "Latency" option. After that's done running, it will have the servers it picked with a check on them. After that, you can look at the one's it picked, and unselect some of them if you want. I did a couple for myself, pick the ones closest to me. After this, make sure to hit the save button at the bottom.

 

However I'm not done with the preference settings yet... After getting a small number of servers picked based off latency by it's own testing, and also checking over them yourself. I wouldn't have no more then 8-10 servers selected.

 

Now scroll down a little, and you will see "No. download streams" with a box beside it. You can pick between 1-32. I honestly think for most of us "4" will be enough. That will be the amount of servers it will use while testing. Do the same for the upload side, which is right below it.

 

For awhile I was using 16+ for each test I done, and I now don't recommend using that high of a number of servers for each test. Reason being, your overall result's will depend on all the servers used during that test. For me, I'm on the east coast of the US, and when I'm doing one of these test using 16+ servers. I'm likely using several servers on the west coast of the US. With a mixature of servers all over, there's so many things that can affect your routing from one minute to the next.

 

So my final opinion on this, keep the number of servers used for the test low, and use servers close to you. Because this to me will give you more accurate result's, when it comes to the bufferbloat part of this test. Finally make sure after any preference setting changes, you make sure to click the "save" button at the bottom.

 

 

 

5. After doing the stuff above, I have been seeing more steady result's with the dslreports.com speedtest, in terms of bufferbloat, and my overall results. I don't feel the current default settings used for their speedtest, is optimized very well for testing, and getting the most accurate results. Ultimately I think by default they're running the test using to many servers, which can be from far away. With that, comes a lot of possible routing issue's. Some of them you have no control over, and isn't an issue with your connection. However whatever each server is reporting back, ultimately is included at the end of the test, and will factor into your final result.

 

 

 

Thanks for the post buddy, just jumped from the R1 over to the XR yesterday and you saved me and likley the duma guys some headache. Every new owner of the XR even those like myself who thought they wouldn’t need help coming from the R1 should read this and the sticky above this one. As always this forum is half of what makes netduma so good. Really liking the XR so far, loved the R1, but couldn’t help but see it’s age every time I looked up at it or attempted to use WiFi more than a few arms lengths away.

[Blameless]

Hi and thank you for your info! But I did all this and I still get an A A A!! Like you said I'm only using 3 servers to ping! I'm using xr500 gaming router! And I also notice that if I dont check the share access bandwidth box I dont get good results...so I enable the share access bandwidth and I get A A A!! Help thank you!

[Netduma Admin]

Hi and thank you for your info! But I did all this and I still get an A A A!! Like you said I'm only using 3 servers to ping! I'm using xr500 gaming router! And I also notice that if I dont check the share access bandwidth box I dont get good results...so I enable the share access bandwidth and I get A A A!! Help thank you!

 

Hi Blameless - you should definitely keep share excess enabled otherwise you will be limiting your connection.

 

Could you try this please:

1. Set Anti-Bufferbloat to Always mode
2. Reduce the sliders to 50% for both download and upload
3. Open the Anti-Bufferbloat advanced options and make sure you have entered the correct total speeds for your bandwidth
4. Disable IPv6 if you enabled it (in >Settings >Advanced Settings >IPv6

[Blameless]

Hi and thank you for the reply! I run my speed test again the sliders to 70 percent ...and I started playing around with DL&UL on the bandwidth allocation for my xbox one x!! And I finally got an A+ BUFFERBLOAT! DL4% 12.03MBPS&UL 0.4% 0.57MBPS!!