QoS Mysteries

[shelleyevans]

I do recommend it tho

Oh yes, me too!! QoS will definitely reduce bufferbloat-- maybe not perfectly, but reliably, as you can see from my graphs above. But the script can very easily create customized containers for your traffic. For me this makes the difference between VOIP working and not working, even when QoS is enabled.

 

[shelleyevans]

I have a question, if anybody following this thread has some time. I recently updated firmware on my AiMesh system, and reinstalled the script, which got me checking bufferbloat again. Frustrated by my difficulty getting good bufferbloat scores I tried lowering my download bandwidth to 275mpbs. This is 25% of my ISP rated speed of 1000mbps. To my shock, I suddenly am getting bufflerbloat scores of A and sometimes A+. More interesting to me, the measured download speeds on DSLReports speedtest are always very steady-- about 10 mbps lower than my cap, ~265mbps. This creates a "flat" graph, exactly the kind that @FreshJR recommends.


When I set my download bandwidth higher (anywhere from 300mbps to 750mbps), two things happen, repeatedly-- 1) the measured download speeds fluctuate pretty widely. 2) On my first speed test I usually get a "B" for bufferbloat, and then on the second test I might get an "A."



My conclusion is that even though my ISP says I'm getting 1000mbps download, from moment to moment I might be getting anything from 275 - 975mbps. If I want my bufferbloat to be good, I have to set my download bandwidth to the lowest measured speed. Does this sound right?

If so, my feelings are mixed: I'm delighted to now have predictable DSL measurements, and good buffer bloat scores. But it's frustrating to pay for gigabit speeds and only be able to use so little of it. Thoughts?

 

[OzarkEdge]

I have a question, if anybody following this thread has some time. I recently updated firmware on my AiMesh system, and reinstalled the script, which got me checking bufferbloat again. Frustrated by my difficulty getting good bufferbloat scores I tried lowering my download bandwidth to 275mpbs. This is 25% of my ISP rated speed of 1000mbps. To my shock, I suddenly am getting bufflerbloat scores of A and sometimes A+. More interesting to me, the measured download speeds on DSL speedtest are always very steady-- about 10 mbps lower than my cap, ~265mbps. This creates a "flat" graph, exactly the kind that @FreshJR recommends.
View attachment 14548

When I set my download bandwidth higher (anywhere from 300mbps to 750mbps), two things happen, repeatedly-- 1) the measured download speeds fluctuate pretty widely. 2) On my first speed test I usually get a "B" for bufferbloat, and then on the second test I might get an "A."

View attachment 14549

My conclusion is that even though my ISP says I'm getting 1000mbps download, from moment to moment I might be getting anything from 275 - 975mbps. If I want my bufferbloat to be good, I have to set my download bandwidth to the lowest measured speed. Does this sound right?

If so, my feelings are mixed: I'm delighted to now have predictable DSL measurements, and good buffer bloat scores. But it's frustrating to pay for gigabit speeds and only be able to use so little of it. Thoughts?

Just a few ad-hoc thoughts...

I'm amazed to hear of Gigabit DSL since copper pair is so susceptible to noise and distance.

ASUS firmware 32797 is suppose to fix some long standing QoS defect. It also broke my wireless client connection to my 86U AiMesh node... speed fluctuates wildly. I'm wondering if they really fixed the QoS defect or just made things worse? If the later, then maybe you can't draw a firm conclusion about your QoS... if you are using firmware 32797 or 32799.

OE

 

[shelleyevans]

I'm amazed to hear of Gigabit DSL

Sorry! I should have said "DSLReports" speedtest! I have cable internet. I will correct original post. I am running the latest (32799) firmware, but doing all my tests on a wired connection, wired backhaul, which should (in theory) be stable.

I realize that I might seem to be beating a dead horse with this bufferbloat struggle, but I had a terrible VOIP business call the other day. My impression is that the better my bufferbloat the better my voice quality. POSSIBLY what made my call quality bad was this new firmware. I will see how my calls are with my new set up (275/41). If improved, I will credit QoS. If it's still bad, I will blame the firmware!

 

[OzarkEdge]

Sorry! I should have said "DSLReports" speedtest! I have cable internet. I will correct original post. I am running the latest (32799) firmware, but doing all my tests on a wired connection, wired backhaul, which should (in theory) be stable.

I realize that I might seem to be beating a dead horse with this bufferbloat struggle, but I had a terrible VOIP business call the other day. My impression is that the better my bufferbloat the better my voice quality. POSSIBLY what made my call quality bad was this new firmware. I will see how my calls are with my new set up (275/41). If improved, I will credit QoS. If it's still bad, I will blame the firmware!

My cable ISP speeds are a nominal 100/10 Mbps. Buffer bloat results are never good.

No router QoS settings. WAN\NAT Passthrough\SIP Passthrough is Disabled.

VoIP service is OBi202 ata on VoIP.ms and Zoiper on Android mobiles. No issues. VoIP uses very little bandwidth.

What are your VoIP issues, service, equipment?

OE

 

[shelleyevans]

Hi! I went into detail about this on another thread, but basically I was finding that my VOIP calls, either the ATA's (Obi202 and two Obi1062 phones) or internet calling, would regularly come through with choppy or strange audio. I went on a long journey starting last May to make my internet more stable, and once I installed this script I didn't have any more problems. After installing the new firmware (32799), I had a bad call for the first time in a long time (a month?). So possibly it is the firmware. I won't know until I have another long conference call, which might not be for a week or so. I do have great bufferbloat scores now, since I cut my speed by 3/4!

@FreshJR said earlier in this thread that QoS might not be able to handle gigabit speeds. I wonder if that's why I had to lower the bandwidth so much.

 

[shelleyevans]

WAN\NAT Passthrough\SIP Passthrough is Disabled.

Somehow my voice quality is still fuzzy. Could you clarify the above settings for me? Are they the same as mine, posted below? Thank you! At the moment these are my settings:

 

[OzarkEdge]

Somehow my voice quality is still fuzzy. Could you clarify the above settings for me? Are they the same as mine, posted below? Thank you! At the moment these are my settings:

My only non-default setting is to disable SIP Passthrough. So, all are enabled except the last two.

OE

 

[Vexira]

Just a few ad-hoc thoughts...

I'm amazed to hear of Gigabit DSL since copper pair is so susceptible to noise and distance.

ASUS firmware 32797 is suppose to fix some long standing QoS defect. It also broke my wireless client connection to my 86U AiMesh node... speed fluctuates wildly. I'm wondering if they really fixed the QoS defect or just made things worse? If the later, then maybe you can't draw a firm conclusion about your QoS... if you are using firmware 32797 or 32799.

OE

It does exist look up g. Fast and other vdsl based technology.