QoS Effects

  • ATTENTION! As of November 1, 2020, you are not able to reply to threads 6 months after the thread is opened if there are more than 500 posts in the thread.
    Threads will not be locked, so posts may still be edited by their authors.
    Just start a new thread on the topic to post if you get an error message when trying to reply to a thread.

TheLyppardMan

Very Senior Member
Having had my network virtually come to a standstill the other day when my Synology Diskstation was uploading data to the cloud, I thought it might be a good idea to try the various types of QoS to see which one works best for me. I'm also keeping an eye on the spdMerlin speeds tests to see what effect each type of QoS has. So far, I have just tried ASUS Adaptive QoS, FlexQoS and Traditional QoS, which I'm currently running and which so far, seems to be working just fine. Obviously, I need to test each method for a while before coming to any decision. Q: Can anyone advise me how to best stress test the system to see how it copes under load?

For interest, these are the effects on the speed test results so far and the download/upload settings I am using...
 

Attachments

Treadler

Very Senior Member
Having had my network virtually come to a standstill the other day when my Synology Diskstation was uploading data to the cloud, I thought it might be a good idea to try the various types of QoS to see which one works best for me. I'm also keeping an eye on the spdMerlin speeds tests to see what effect each type of QoS has. So far, I have just tried ASUS Adaptive QoS, FlexQoS and Traditional QoS, which I'm currently running and which so far, seems to be working just fine. Obviously, I need to test each method for a while before coming to any decision. Q: Can anyone advise me how to best stress test the system to see how it copes under load?

For interest, these are the effects on the speed test results so far and the download/upload settings I am using...
Search the net for ‘bufferbloat’ as a starting point.
An interesting topic, that can be as big as you want it to be.:)

Bufferbloat, & other interesting stuff can be tested here;

 

sbsnb

Senior Member
What I usually do is first get an idea of the real up/down speeds the router sees from the ISP (not their advertised speeds). Some connections, like fiber, can be rock solid at a given rate. Others, like cable internet, can be all over the place depending on neighborhood load. I find spdMerlin to be good for establishing a connection's lowest up/down speeds in a typical day. Once I have a good idea of the lowest speeds seen by the router I'll set the QoS up/down limits to about 90 - 95% of those speeds.

After setting the QoS speeds I'll then fire up the continuous ping in "Network Tools" from the router's web ui. I let that go for a minute or two to establish a baseline and the fire up something that consumes a nasty amount of bandwidth on a wired client (like bittorrent a linux distro). Then I watch the ping times. If they don't change while downloading at full speed I call it a day. If the ping times still spike I try lowering the QoS speed limits some and try again.
 

TheLyppardMan

Very Senior Member
This sounds like a good plan to me. I'll bookmark this thread so I can refer to it when needed. Thanks folks.
 

sbsnb

Senior Member
If it's for family/customers who may not like the idea of a speed limit much lower than their ISP's advertised speed, I've had success explaining it with a traffic simile. I tell them a simplified something like:

Your internet connection services the entire neighborhood, and your speed is variable depending on how many people in the neighborhood are using their internet and how much of it they're using. Internet traffic can be like highway traffic in that there are traffic jams when there is more traffic than the highway can handle efficiently. If we set your network's speed limit too high you will be able to go faster when traffic congestion improves, but you'll also have to slam on the brakes at the next slowdown. It's that slamming on the brakes that can make everything in the house grind to a virtual halt - like a 'smack up in the back up.' To make your network operate smoothly and avoid the instability we set a speed limit that is around the slowest speed in that traffic jam so that your traffic continues at a stable pace and error-free at almost all times.
 

heysoundude

Very Senior Member
If it's for family/customers who may not like the idea of a speed limit much lower than their ISP's advertised speed, I've had success explaining it with a traffic simile. I tell them a simplified something like:
Then what's the point of paying for the speeds of cable (when you can reach them) if you only have to slow things down for reliability of connection?

I've never had anything but DSL over copper phone lines; 50/10 is good and stable for me thank you very much. If someone brings a direct fibre link to me, maybe I might feel differently...
 

sbsnb

Senior Member
It depends on the provider. I've dealt with Spectrum the most, and their impacted times seem to slow download speeds to about half of their advertised speeds. So you buy 400 mbps service, but you can still reliably pull 200 mbps. What bugs me about cable is the slow upload speeds (although still faster than your situation). It's horrible when you're doing VPN work.
 

heysoundude

Very Senior Member
It depends on the provider. I've dealt with Spectrum the most, and their impacted times seem to slow download speeds to about half of their advertised speeds. So you buy 400 mbps service, but you can still reliably pull 200 mbps. What bugs me about cable is the slow upload speeds (although still faster than your situation). It's horrible when you're doing VPN work.
I could (if I were to upgrade my mobile device and plan) -using DualWAN and tethering- achieve your DL (or higher) and >triple my UL because I live ~200yds from one of my provider's tx sites and ~400yds from their roaming partner's (whose speeds are significantly higher than my provider's)...at the expense of at least tripling my ping times (the redundancy is comforting, though)
what I'd love to know/understand: if under DualWAN, is QoS applied to the aggregate speed, or would it need to be configured prior to DualWAN (meaning an instance on each interface/service)?
 

TheLyppardMan

Very Senior Member
Search the net for ‘bufferbloat’ as a starting point.
An interesting topic, that can be as big as you want it to be.:)

Bufferbloat, & other interesting stuff can be tested here;

Just ran a test on this web site and got the following results:-
Screenshot - 07_09_2020 , 17_13_45.jpg


I tried it about an hour ago as well and both were rated A+, so I may be getting close to the ideal settings, but I'll carry on testing for a while to see if I need to make any more tweaks.
 

TheLyppardMan

Very Senior Member
Just tried it again; not much load on the network at the moment though - just web surfing on my laptop + youtube/web surfing on my daughter's devices + streaming from two security cameras to an Huawei Android tablet. Q; What do the red range bars signify exactly?
Screenshot - 07_09_2020 , 20_00_39.jpg
 

TheLyppardMan

Very Senior Member
What I usually do is first get an idea of the real up/down speeds the router sees from the ISP (not their advertised speeds). Some connections, like fiber, can be rock solid at a given rate. Others, like cable internet, can be all over the place depending on neighborhood load. I find spdMerlin to be good for establishing a connection's lowest up/down speeds in a typical day. Once I have a good idea of the lowest speeds seen by the router I'll set the QoS up/down limits to about 90 - 95% of those speeds.

After setting the QoS speeds I'll then fire up the continuous ping in "Network Tools" from the router's web ui. I let that go for a minute or two to establish a baseline and the fire up something that consumes a nasty amount of bandwidth on a wired client (like bittorrent a linux distro). Then I watch the ping times. If they don't change while downloading at full speed I call it a day. If the ping times still spike I try lowering the QoS speed limits some and try again.
I could also try leaving Multiping running for a day or so and see what results it comes up with (I've just reinstalled it - see screenshot).Screenshot - 07_09_2020 , 20_34_15.jpg
 

TheLyppardMan

Very Senior Member
...After setting the QoS speeds I'll then fire up the continuous ping in "Network Tools" from the router's web ui. I let that go for a minute or two to establish a baseline and the fire up something that consumes a nasty amount of bandwidth on a wired client (like bittorrent a linux distro). Then I watch the ping times. If they don't change while downloading at full speed I call it a day. If the ping times still spike I try lowering the QoS speed limits some and try again.
I think some file uploads from my NAS to iDrive should suffice for that, as my Synology Diskstation has a wired connection to the router. I'll try it after work tomorrow and see what happens.
 

Treadler

Very Senior Member
Just ran a test on this web site and got the following results:-
View attachment 26057

I tried it about an hour ago as well and both were rated A+, so I may be getting close to the ideal settings, but I'll carry on testing for a while to see if I need to make any more tweaks.
Yup, you will never get a perfect result every time. You’re just looking for the settings that give you the most ‘best’ results, ‘most of the time’.

The red lines give you indications of the min/max bufferbloat levels seen during a test. As opposed to the solid colour graph which shows the average over the whole test.
Clicking on the solid colour will give the you detail......
 

TheLyppardMan

Very Senior Member
Yup, you will never get a perfect result every time. You’re just looking for the settings that give you the most ‘best’ results, ‘most of the time’.

The red lines give you indications of the min/max bufferbloat levels seen during a test. As opposed to the solid colour graph which shows the average over the whole test.
Clicking on the solid colour will give the you detail......
OK, got it. Thanks for that.
 

Wade Coxon

Regular Contributor
Please forgive the potentially private question, but why is your connection more or less 100% utilised 24/7? I am assuming this is some sort of bulk upload/download that is happening? If it is bulk downloads, then I would strongly recommend throttling download/upload at the client side to no more than 80% of your measured link speed.
If your link is saturated all the time, then your day to day usage experience is likely to be affected in some way.
 

BikeHelmet

Occasional Visitor
What I usually do is first get an idea of the real up/down speeds the router sees from the ISP (not their advertised speeds). Some connections, like fiber, can be rock solid at a given rate. Others, like cable internet, can be all over the place depending on neighborhood load. I find spdMerlin to be good for establishing a connection's lowest up/down speeds in a typical day. Once I have a good idea of the lowest speeds seen by the router I'll set the QoS up/down limits to about 90 - 95% of those speeds.

After setting the QoS speeds I'll then fire up the continuous ping in "Network Tools" from the router's web ui. I let that go for a minute or two to establish a baseline and the fire up something that consumes a nasty amount of bandwidth on a wired client (like bittorrent a linux distro). Then I watch the ping times. If they don't change while downloading at full speed I call it a day. If the ping times still spike I try lowering the QoS speed limits some and try again.
Oh boy...

Fiber: 940/940

Cable: 150/15

Equipment quality has a big impact.

The red lines give you indications of the min/max bufferbloat levels seen during a test. As opposed to the solid colour graph which shows the average over the whole test.
Clicking on the solid colour will give the you detail......
The red lines are ping times in absolute ms values. Bufferbloat is a computed number that shows the difference in time between when packets were expected and when they actually arrived. It's a guestimate, but a pretty accurate one, and 0ms bufferbloat would be relative to your lowest stable ping time. Wide red bars indicate your connection is volatile and might suck for VOIP/gaming, while tight ones indicate that it'll probably be perfectly fine.

Please forgive the potentially private question, but why is your connection more or less 100% utilised 24/7? I am assuming this is some sort of bulk upload/download that is happening? If it is bulk downloads, then I would strongly recommend throttling download/upload at the client side to no more than 80% of your measured link speed.
If your link is saturated all the time, then your day to day usage experience is likely to be affected in some way.
Isn't that the automated speed tester? His connection isn't in use - every 30 minutes it runs a speed test and graphs the results. To get those stable speeds in the speedtests, his connection must be idle/unused when the test is running. (So opposite of what you thought.)
 

sbsnb

Senior Member
Oh boy...

Fiber: 940/940

Cable: 150/15
Sorry to hijack the thread a little, but how did you get bufferbloat test to work on dslreports? They've been broken for me for years, just showing a blank area where the bufferbloat score should appear.

2020-09-08 05_57_10-11ms 86.3_130.3.png
 

TheLyppardMan

Very Senior Member
My son and daughter have been living a pretty nocturnal lifestyle while the schools/colleges have been closed due to coronvirus, hence the more or less continuous use of the network. Also, my Synology NAS backs up my data to iDrive during the small hours.

My latest ping results from Multiping results (on my laptop, which I left running overnight to see what effect the iDrive uploads would have) and spdMerlin results are attached.
 

Attachments

sbsnb

Senior Member
The ping graph is too tiny to see here, but from what I can see it looks pretty good. What would be helpful for assessing would be to know the start and stop times of the iDrive uploads and the speed of those uploads. Are they saturating your upload speed?
 

Similar threads

Top