spdMerlin spdMerlin - v4.x - speedtest and bandwidth monitoring

[MvW]

So is it possible that for every Speedtest time it automatically do multiple Speedtests concurrently?

Wouldn't 3 speedtests concurrently affect your maximum achievable speed? I assume each speedtest tries to utilize the bandwidth available as optimal as possible, hence running 3 speedtests at the same time will most likely only cause congestion and therefore bring the maximum achievable speeds down, hence you won't have a realistic measurement anymore. I'm trying to see what you would like to achieve here...

 

[Matthew Patrick]

Wouldn't 3 speedtests concurrently affect your maximum achievable speed? I assume each speedtest tries to utilize the bandwidth available as optimal as possible, hence running 3 speedtests at the same time will most likely only cause congestion and therefore bring the maximum achievable speeds down, hence you won't have a realistic measurement anymore. I'm trying to see what you would like to achieve here...

I'm sorry. English is not my primary language. What I meant was sequentially

 

[Jack Yaz]

sequential tests would be feasible, but not with a different server for each test. coding that (for both webui and cli) would be a nightmare

 

[MvW]

I'm sorry. English is not my primary language. What I meant was sequentially

No worries. Neither is it for me. I actually had to Google 'concurrently' to verify if you actually wrote what I thought you meant. Sequentially makes much more sense.

 

[Wistuplu]

Fun.

The multiple target of Matthew Patrick was interesting, and a bit wild.
I would have used it to target one server inside my ISP network (as this is the feedback to my ISP prefers), and one server outside my ISP network (as this is what's most relevant to me).

Another even wilder idea:
Define a threshold that indicates an incident on the network. There would be two different frequencies of execution of the tests. If the latest test was under the threshold, the exceptional frequency applies (typically defined as more frequent). If the latest test was over the threshold, it switches back to the normal frequency (typically less frequent). It would be used to get improved visibility on the end of an incident.

NB: And the next step is having two thresholds. One to enter the "exceptional" context (e.g. 230Mbps) and one to reenter the normal context (e.g. 280 Mbps), considering e.g. a line with a marketed 400Mbps, and a "minimal" of 230Mbps.

And the beauty of it is that Jack Yaz was nice enough to make the code available on Github and with a GPL license. So nothing stops us from implementing it ourselves! (if we had the skills ;-) )

Enjoy the weekend
Take care

W.

 

[Matthew Patrick]

No worries. Neither is it for me. I actually had to Google 'concurrently' to verify if you actually wrote what I thought you meant. Sequentially makes much more sense.

Yeh today was not my day I kinda typed that message quick so I didn't double check everything

sequential tests would be feasible, but not with a different server for each test. coding that (for both webui and cli) would be a nightmare

No problem!! Just a wild idea

Fun.

The multiple target of Matthew Patrick was interesting, and a bit wild.
I would have used it to target one server inside my ISP network (as this is the feedback to my ISP prefers), and one server outside my ISP network (as this is what's most relevant to me).

Yep. I'd prob do that too.. so one say with auto server or manual. Then the other one is fixed to my ISPs speedtest server. So they can't deny things when I send over the csv and graphs with reasoning like. Oh we need the Speedtests server to be our server etc

 

[MvW]

@Jack Yaz, I'm not sure whether this is related to my posts in the CakeQos-Merlin thread or to the connmon development version, but today I'm experiencing issues with speedtests again. I've seen speed decreasing throughout the day. Yesterday my ISP raised my bandwidth from 250/25 to 300/30 so the only thing I've done after an hour and a half is temporarily set the number used for calculating average speeds to 3 and then around midnight I set it back to it's default value of 10. Then the speeds measurements starting dropping, and I'm now on average of about 60% of the available bandwidth. Exclude from QoS is selected but greyed out. I've set the values on the QoS tab on 100%, 30 up, 300 down before running the speedtests below and the Ookla Speedtest and a speedtest through spdMerlin:

It looks like the issue I previous had, but I find it odd I have two, three spikes where a much higher speed is achieved and when I manually override the values, I get nearly my complete bandwidth back again. Any idea what's going on?

 

[Jack Yaz]

@Jack Yaz, I'm not sure whether this is related to my posts in the CakeQos-Merlin thread or to the connmon development version, but today I'm experiencing issues with speedtests again. I've seen speed decreasing throughout the day. Yesterday my ISP raised my bandwidth from 250/25 to 300/30 so the only thing I've done after an hour and a half is temporarily set the number used for calculating average speeds to 3 and then around midnight I set it back to it's default value of 10. Then the speeds measurements starting dropping, and I'm now on average of about 60% of the available bandwidth. Exclude from QoS is selected but greyed out. I've set the values on the QoS tab on 100%, 30 up, 300 down before running the speedtests below and the Ookla Speedtest and a speedtest through spdMerlin:

View attachment 33182
View attachment 33181
View attachment 33183

It looks like the issue I previous had, but I find it odd I have two, three spikes where a much higher speed is achieved and when I manually override the values, I get nearly my complete bandwidth back again. Any idea what's going on?

if it was autobw/qos you'd see a continuous downward trend in speeds. this is more likely the ookla servers, try picking a specific server and see how that one performs

 

[MvW]

if it was autobw/qos you'd see a continuous downward trend in speeds. this is more likely the ookla servers, try picking a specific server and see how that one performs

After manually editing bandwidth (which was already set back again to about 60%) I get my full speed at the nearest Ookla speedserver:

2021-04-14 21:05:37    301.98    31.0    9.05    0.95    0.0    404.2    15.5    Speedtest result URL

Speedtest by Ookla - The Global Broadband Speed Test

Use Speedtest on all your devices with our free desktop and mobile apps.

www.speedtest.net

Besides the last time with AutoBW, my speeds are always stable:

And now today they suddenly collapse, but give the expected results when I override the values calculated by AutoBW. I don't mean to be disrepectful, but that seems highly unlikely, to be honest. I'm one of many, many customers of the largest ISP in my country and if something went wrong with the countrywide upgrade of the speed limits, which started yesterday at noon officially, there would be a massive amount of complaints, but there aren't any on their forums. Apparently only my speed is suddenly fluctuating. Both up- and downstream as you can see. Something isn't right, but I don't have the knowledge to tell you where it goes wrong.

 

[Jack Yaz]

After manually editing bandwidth (which was already set back again to about 60%) I get my full speed at the nearest Ookla speedserver:

2021-04-14 21:05:37    301.98    31.0    9.05    0.95    0.0    404.2    15.5    Speedtest result URL

Speedtest by Ookla - The Global Broadband Speed Test

Use Speedtest on all your devices with our free desktop and mobile apps.

www.speedtest.net

Besides the last time with AutoBW, my speeds are always stable:

View attachment 33184

And now today they suddenly collapse, but give the expected results when I override the values calculated by AutoBW. I don't mean to be disrepectful, but that seems highly unlikely, to be honest. I'm one of many, many customers of the largest ISP in my country and if something went wrong with the countrywide upgrade of the speed limits, which started yesterday at noon officially, there would be a massive amount of complaints, but there aren't any on their forums. Apparently only my speed is suddenly fluctuating. Both up- and downstream as you can see. Something isn't right, but I don't have the knowledge to tell you where it goes wrong.

i'm a customer of Virgin Media and if there's a problem with the CMTS in my area, my speeds go from 550 down to < 50 down, with upload dropping from 35 up to 15-20 up. faults happen, and don't always necessarily affect everyone the same

are you using a fixed server, or automatic?

autobw uses the last 10 (or3, in your case) speedtest results to calculate the average, and should print a message to syslog whenever an adjustment is made. do you see corresponding messages in your system log? they'll be along the lines of "Setting QoS download/upload speed..."
if you can share syslog with these in i can work back through the (mis)calculations if thats whats happening

 

[MvW]

autobw uses the last 10 (or3, in your case) speedtest results to calculate the average, and should print a message to syslog whenever an adjustment is made. do you see corresponding messages in your system log? they'll be along the lines of "Setting QoS download/upload speed..."
if you can share syslog with these in i can work back through the (mis)calculations if thats whats happening

I also use 10 measurements. I only used 3 once, over 3 measurements, to have the QoS reflect the new, upgraded bandwidth. After that, I've set it back to 10 and saved it, so it has been at ten since midnight (a little 22 hours ago).

Yesterday at 15:20 I rebooted my ISP's Compal modem to enforce to speed upgrade (otherwise they'll reboot your modem remotely at night).

You can see a manual speedtest initiated at 15:23 after my ISP Modem got back online and the router showed "Connected", to confirm the new speeds were deployed.

Here's spdmerlin.log: https://privatebin.net/?8f21fba9bf2c2b4e#8rexSgBeJPe7MYP8Nj83zvLQHgpcTBtoVgUWhkoFCZyR

 

[Jack Yaz]

I also use 10 measurements. I only used 3 once, over 3 measurements, to have the QoS reflect the new, upgraded bandwidth. After that, I've set it back to 10 and saved it, so it has been at ten since midnight (a little 22 hours ago).

Yesterday at 15:20 I rebooted my ISP's Compal modem to enforce to speed upgrade (otherwise they'll reboot your modem remotely at night).

View attachment 33186

You can see a manual speedtest initiated at 15:23 after my ISP Modem got back online and the router showed "Connected", to confirm the new speeds were deployed.

Here's spdmerlin.log: https://privatebin.net/?8f21fba9bf2c2b4e#8rexSgBeJPe7MYP8Nj83zvLQHgpcTBtoVgUWhkoFCZyR

Starting speedtest using using auto-selected server for WAN interface

using an auto selected server means your tests could be using different servers each time, and its well known some perform better than others. i recommend testing against specific servers and then opting to use the "best" one as preferred, and then keep an eye on it again

 

[MvW]

Starting speedtest using using auto-selected server for WAN interface

using an auto selected server means your tests could be using different servers each time, and its well known some perform better than others. i recommend testing against specific servers and then opting to use the "best" one as preferred, and then keep an eye on it again

Okay, will do as you requested. I've got a reliable server at 6km distance which is the Ookla speedtest in the UI auto-selects 99% of the time, so I'm fairly certain the Ookla CLI speedtest client does the same. Out of curiousity: did you look at the 30 day graph I posted here? Am I missing something here?

 

[MvW]

I'm using a fixed server now (or actually, two: one for WAN and another for VPNC1 because of different endpoints). CakeQoS is disabled as I'm having issues with it and AutoBW has been disabled too. I've switched to Adaptive QoS and have manually entered 90% of up- and download bandwidth this morning at 08:40 (CEST) and my speed is back at around 300Mbit Down and 30 Mbit up.

 

[MvW]

@Jack Yaz, all seems back to normal now. I've removed all the scripts I barely use, set two fixed, reliable speedtest servers and the results are very consistent now. Cake has been re-enabled and all is well. I don't have a clue what happened, maybe it's time to start fresh someday soon.

One thing I was wondering about: why takes it longer than a reboot to save changes to one of the variables of spdMerlin. If I, for instance, only change the percentage of bandwidth I want to make available for QoS, and I change it from 95% to 90%, it takes over a minute to save this value (which on a RT-AC86U takes less than a minute). Are there so many values rewritten or does it take a lot of processes to be restarted?

 

[Jack Yaz]

@Jack Yaz, all seems back to normal now. I've removed all the scripts I barely use, set two fixed, reliable speedtest servers and the results are very consistent now. Cake has been re-enabled and all is well. I don't have a clue what happened, maybe it's time to start fresh someday soon.

One thing I was wondering about: why takes it longer than a reboot to save changes to one of the variables of spdMerlin. If I, for instance, only change the percentage of bandwidth I want to make available for QoS, and I change it from 95% to 90%, it takes over a minute to save this value (which on a RT-AC86U takes less than a minute). Are there so many values rewritten or does it take a lot of processes to be restarted?

The timeout for the page reload should be the same for all models, as its a javascript variable fed to the "loading" screen. i may have set it a bit on the high side out of caution to make sure the backend service calls were completed, I can take a look

 

[Stephen Harrington]

i may have set it a bit on the high side out of caution to make sure the backend service calls were completed,

@Jack Yaz, yes it DOES seem to be pretty long.
The first time I changed something I figured it must have been doing a full reboot of the router ...
In fact now with my fancy new AX86U I think a reboot would actually be shorter

Best wishes for your dog from us!

 

[Jack Yaz]

@Jack Yaz, yes it DOES seem to be pretty long.
The first time I changed something I figured it must have been doing a full reboot of the router ...
In fact now with my fancy new AX86U I think a reboot would actually be shorter

Best wishes for your dog from us!

ah, i see the problem. where i tidied up some lines, the timeout variable is being set to 10s... AFTER the page is submitted, oops

 

[Jack Yaz]

Hotfix for settings saved pushed out, no version change

 

[MvW]

Hotfix for settings saved pushed out, no version change

Works like a charm, the difference is clearly noticable. Thanks for the quick hotfix.