Recommended default QoS settings for a connection with fluctuating D/L Bandwidths

[Moisala]

Hey,

I'm unfortunate enough to live in a location where I can't get a proper cable or broadband connection. Only a very slow ADSL connection is available.

I then decided to choose the 4G (up to 100Mbit/s connection)  + 4G router + public IP option  the ISP offered me.

I was pleasantly surprised to see that I could play online. As far as I know, the latency with 4G is almost always (or always) higher than with a cable or a DSL connection. Anyway, I can play competitive with my 4G connection.

Now to the problem/question:

My download/upload bandwidths aren't stable and are hugely affected by the usage of the 4G towers nearby. I've been testing the connection with Speedtest and during the rush hours the download is around 1/3 of the bandwidth I get during the quiet hours. The download bandwidth fluctuates between ~15 to 50 Mbps and the upload fluctuates between 25-50 Mbps. The low end speeds obviously happen during the rush hours and the high ones when the variables are on my side. 

What Download / Upload Bandwidth numbers should I use on default?

I'm guessing the Anti-Bufferbloat (both set to 70%) doesn't work properly if the values used in D/L Bandwidth are set too far from the reality. Let's say I use the speed figures I get during the quiet hours (50/50 Mbps) and use 70%/70%. 

Then I start playing during the rush hours using those settings and values. The reality probably is, that the D/L Bandwidths are much lower (let's say 18/25 Mbps) at that point and I'm guessing the Anti-Bufferbloat is designed to work based on the speed values I've set. 

This situation can't be optimal, right?  

I know I could run the Speedtest every time before I start playing and use those values but I'd rather not if there's a less wearing option available. 

Any ideas/solutions?

[BIG__DOG]

7 hours ago, Moisala said:

Hey,

I'm unfortunate enough to live in a location where I can't get a proper cable or broadband connection. Only a very slow ADSL connection is available.

I then decided to choose the 4G (up to 100Mbit/s connection)  + 4G router + public IP option  the ISP offered me.

I was pleasantly surprised to see that I could play online. As far as I know, the latency with 4G is almost always (or always) higher than with a cable or a DSL connection. Anyway, I can play competitive with my 4G connection.

Now to the problem/question:

My download/upload bandwidths aren't stable and are hugely affected by the usage of the 4G towers nearby. I've been testing the connection with Speedtest and during the rush hours the download is around 1/3 of the bandwidth I get during the quiet hours. The download bandwidth fluctuates between ~15 to 50 Mbps and the upload fluctuates between 25-50 Mbps. The low end speeds obviously happen during the rush hours and the high ones when the variables are on my side. 

What Download / Upload Bandwidth numbers should I use on default?

I'm guessing the Anti-Bufferbloat (both set to 70%) doesn't work properly if the values used in D/L Bandwidth are set too far from the reality. Let's say I use the speed figures I get during the quiet hours (50/50 Mbps) and use 70%/70%. 

Then I start playing during the rush hours using those settings and values. The reality probably is, that the D/L Bandwidths are much lower (let's say 18/25 Mbps) at that point and I'm guessing the Anti-Bufferbloat is designed to work based on the speed values I've set. 

This situation can't be optimal, right?  

I know I could run the Speedtest every time before I start playing and use those values but I'd rather not if there's a less wearing option available. 

Any ideas/solutions?

Hi and welcome to the forum.

Sorry to hear that you are forced to use a $G connection as apart from it being wireless it is also connected to the mobile network and therefore your gaming data is competing alongside normal mobile data. I would input the average speeds you get into the QOS at the time that you game mainly. If it is in the evening then input those speeds and set sliders to 707/70 and to "Always" once gaming has finished totally disable QOS and you can use the maximum bandwidth possible for you without inputting fresh ul/dl data into the QOS.

[Moisala]

Hey BIG_DOG,

 

Thanks for the reply. Your suggestions sound logical. I'll try them out.

 

I haven't actually seen much difference when playing, when I compare my current 4G connection compared to the previous 350M DSL-connection I used not too long ago in my previous living location. I guess I'm lucky in that aspect. 

Though I fear that the situation could change anytime as 4G network and the connection I get has many other variables besides the the amount of users using the network.  Even things like change in the weather or the ISP rotating the antenna of the 4G base station 90 degrees the other way could have a big effect on the quality of my connection.  At least right now the situation is bearable.

 

[BIG__DOG]

5 minutes ago, Moisala said:

Hey BIG_DOG,

 

Thanks for the reply. Your suggestions sound logical. I'll try them out.

 

I haven't actually seen much difference when playing, when I compare my current 4G connection compared to the previous 350M DSL-connection I used not too long ago in my previous living location. I guess I'm lucky in that aspect. 

Though I fear that the situation could change anytime as 4G network and the connection I get has many other variables besides the the amount of users using the network.  Even things like change in the weather or the ISP rotating the antenna of the 4G base station 90 degrees the other way could have a big effect on the quality of my connection.  At least right now the situation is bearable.

 

I couldn't imagine gaming on a 4G network so I sympathise with your situation. There are far more variables outside your controle which forces you to accept what you have however reluctant you may feel. Hopefully a more suitable wired alternative will become available to you sooner rather than later.

[Netduma Fraser]

Let us know how BDs advice helps and if not we can try some other things

[RL317]

Hey Moisala, and welcome! I think I've just conjured up an idea for you to try. 

 

On the old R1 firmware there was a "bug" of sorts where, if you changed the speeds you put into "set bandwidth" but kept your congestion control sliders at 100%, you'd get the max coming through on your line regardless. I found this to be a useful quirk in the system as you could input certain bandwidth values for the purpose of testing and then just move the sliders when you wanted those values to kick in, but getting back to max was as simple as sliding it back up. With the addition of the "when high prio traffic detected" option we can go one better on DumaOS... 

 

I just tried testing this on DumaOS with the high prio traffic mode (and nothing setup in traffic prio) and it works! 

 

So now all we need to set up something decent is a solid minimum estimate for the speeds you're getting. Is it safe to say you never really drop below 15Mb on either the downstream or upstream? If so you could do what I've shown in my screenshot - set yours to 15/15 and drop the QoS sliders to somewhere around 70%. This should be more than sufficient as gaming barely uses more than a few hundred Kb at once. Then click the "when high priority traffic detected" option, and set up traffic prio for your gaming device and test it out with a game or two. Every time you play your fluctuating 50/50Mb~ or whatever is coming through at best will be capped at a stable 15Mb until you get off. No need to turn QoS off after or mess around with your sliders either. 

 

Give it a try, let us know how it goes... and good luck

[Moisala]

Hey everyone,

I'm grateful for your helpfulness!

I like @lllRL 's solution the best as it's so easy and stress-free. I have to test which solution is the best and of course I hope it's the solution for the lazy one 😅

 

 

If I understand at all how this router and these settings work, I think I should maybe drop both download/upload to as low as 10Mb or so just to be safe -- to get as stable gaming connection I could possibly get?  As the lowest values have I've gotten have been below 15 Mb (both D&U).

I need my R1 router + DumaOS just 2-3 times a week for competitive gaming. That's when it matters and that's why I bought this router. And during those time windows my wife's Netflix experience isn't that important and it's ok if it suffers a bit while I play -- as long as she can use streaming services and isn't given 0% of the bandwidth.

So maybe I should heavily limit the bandwidth just for those sessions. I guess I just need the find the lowest safe value for both D/U. The thing is though, that the average values seem to be much higher than 14Mb. Those dips seem to be rare occations. So is it stupid to limit the values to let's say 10Mb if the average values are much higher? 

I haven't had time to test the connection properly on evenings while I usually play. I'm testing right now and after running the Speedtest (same server) the results are:

Download:

Upload:

 

After 18 tests, the average download seems to be close to 60Mb and upload to 30 Mb. The ping has been between 31-33 in every test.

This is the situation right now but it's mid-day. The values are lower during my gaming time  19-22 CET but even without having been able to test it yet, the averages are much lower than these but still much higher than 15/15.  

Looks like only the upload went below 15 once in 18 tests. Download looks good.

This is all speculation until I test the connection at the time I mainly play. But let's say the average values will then be 35/25 or so and either download or upload value goes below 15 once in every 10 test or so.

What values would you use then? Limit heavily like I'm considering or just use the average values?

 

I'll post new graphs when I get the "rush hour results".

[RL317]

1 hour ago, Moisala said:

Hey everyone,

I'm grateful for your helpfulness!

I like @lllRL 's solution the best as it's so easy and stress-free. I have to test which solution is the best and of course I hope it's the solution for the lazy one 😅

 

 

If I understand at all how this router and these settings work, I think I should maybe drop both download/upload to as low as 10Mb or so just to be safe -- to get as stable gaming connection I could possibly get?  As the lowest values have I've gotten have been below 15 Mb (both D&U).

I need my R1 router + DumaOS just 2-3 times a week for competitive gaming. That's when it matters and that's why I bought this router. And during those time windows my wife's Netflix experience isn't that important and it's ok if it suffers a bit while I play -- as long as she can use streaming services and isn't given 0% of the bandwidth.

So maybe I should heavily limit the bandwidth just for those sessions. I guess I just need the find the lowest safe value for both D/U. The thing is though, that the average values seem to be much higher than 14Mb. Those dips seem to be rare occations. So is it stupid to limit the values to let's say 10Mb if the average values are much higher? 

I haven't had time to test the connection properly on evenings while I usually play. I'm testing right now and after running the Speedtest (same server) the results are:

Download:

Upload:

 

After 18 tests, the average download seems to be close to 60Mb and upload to 30 Mb. The ping has been between 31-33 in every test.

This is the situation right now but it's mid-day. The values are lower during my gaming time  19-22 CET but even without having been able to test it yet, the averages are much lower than these but still much higher than 15/15.  

Looks like only the upload went below 15 once in 18 tests. Download looks good.

This is all speculation until I test the connection at the time I mainly play. But let's say the average values will then be 35/25 or so and either download or upload value goes below 15 once in every 10 test or so.

What values would you use then? Limit heavily like I'm considering or just use the average values?

 

I'll post new graphs when I get the "rush hour results".

This'll definitely be the lazy method (can you guess why I thought of it? 😏) as you can just set and forget. However, considering the unique nature of your setup and the serious fluctuations you'd be wise to consider a "day" of testing. Well, more like an evening as peak time for usage on your network will be the best time to test and discover a reliable minimum set of numbers.

 

10/10Mb would be absolutely fine. If you're not 100% confident the line will maintain more than 15Mb all of the time then this will give you a bit of leg room - a buffer to keep you away from the max you're achieving at worst when it drops. Plus 10/10 would be more than enough for gaming anyway. Actually I've been playing around with similar numbers myself lately with bandwidth allocation rather than the "global" QoS sliders and I've found that streaming TV seems to run just fine with lower amounts of bandwidth available for it to use, so you shouldn't need to cut it too close with Netflix while gaming. 

 

Perhaps another test you could try is capping the overall to around 4-7Mb (on the downstream anyway) and testing Netflix to see if the most demanding of content runs without issue. When you're confident you've found a number that works without an issue, you can add on a couple of megabits to factor in gaming and then you can set up your "safe minimum" and give it a proper test in game 👍

 

The only thing to worry about at this point is finding the absolute low end of what you can expect when it's fluctuating. If you see <15 but >10 at worst then you're gonna be just fine! In fact, even if you were getting a constant 15Mb with no dips below, I'd suggest telling the R1 that you're paying for 15/15 and then just set the sliders to 70% for that buffer which'd give you a stable 10Mb anyway. You want the cap to be just below any drops so that you won't get any iffy moments while playing a game if the overall drops below what you've set. Being conservative is definitely better than being optimistic in this scenario lol

 

Edit: maybe you won't need to test Netflix at all. I just found this:

 

"According to Netflix, the Internet speed you'll need for downloads is as follows: For any streaming at all, you'll need a minimum of 0.5 megabits per second (Mbps), but Netflix recommends 1.5 Mbps. For DVD quality, you will need 3.0 Mbps. For HD quality, you will need 5.0 Mbps."

 

This is good. If you never drop below 10Mb then you're going to have more than enough to play with while keeping the wife happy with the best quality streaming 😂 I'd still suggest doing some evening speed tests to be on the safe side, but I think you could get away with telling the R1 you get 10/10, then dropping the sliders to 65-85% and allocating 5Mb to whichever device is using Netflix. Honestly you could get away with even distribution and leaving share excess enabled, but you may want to play around with those to fine tune your setup. You would more than likely be able to get away with 90% sliders as that's an entire megabit shaved off as a buffer, but again you may find the need to fine tune using those sliders/bandwidth allocation or disabling share excess for consistent distribution. 

 

Hopefully your evening speed tests will provide good news and show no dips below 10Mb whatsoever 😄