spdMerlin spdMerlin - v4.x - speedtest and bandwidth monitoring

[iTyPsIDg]

I think it may have been on before. I'll give it a day or two to see if it happens again. Thanks for the help.

 

[dave14305]

I think it may have been on before. I'll give it a day or two to see if it happens again. Thanks for the help.

What is your Maximum Upload Bandwidth percentage for File Downloads in FlexQoS?

 

[dave14305]

Hi @Jack Yaz, one thing to be mindful of when invoking AutoBW is to make sure the user isn't using Automatic Bandwidth in QoS. If they are, both qos_obw and qos_ibw will equal 0 in nvram. No one really recommends using it, but some people still do. I've been running into this more and more lately.

 

[Jack Yaz]

Hi @Jack Yaz, one thing to be mindful of when invoking AutoBW is to make sure the user isn't using Automatic Bandwidth in QoS. If they are, both qos_obw and qos_ibw will equal 0 in nvram. No one really recommends using it, but some people still do. I've been running into this more and more lately.

maybe i should just force Manual in nvram...

 

[iTyPsIDg]

What is your Maximum Upload Bandwidth percentage for File Downloads in FlexQoS?

I left all of my max values at 100%. I tried to only adjust my minimums.

Spoiler: Bandwidth settings

 

[JGrana]

Warning - OOkla might pick the closest - but not always the fastest

I have been looking at an issue with the 386.1/AX88U speedtest results from both spdMerlin and the built in Asus one - both powered by Ookla.

About a week ago, I noticed very erratic speedtest results. I would see my typical 220-230 Mbit download speeds then see a few 40-50 Mbit... I ran a few more spdMerlin tests and also tried the built in speedtest.
Same result, often it would show 40-60Mbit download - my 11 Mbit upload was typically correct.

I had been running FlexQos. Turned it off - same erratic results. Tried Cake-qos - same thing.

I then started to run some speedtests on clients - Analiti on an NVidia Shield and dslreports on a laptop.
Both consistently showed the correct 220-230Mbit download speeds. Every time I ran them.

I did notice that both spdMerlin and Asus Ookla was using a local university as the server (University of Rochester). The ping times to UofR were great ~ 11msec. (BTW, UofR has been a main internet node in this area for a very long time - pre WWW and the .com/edu etc. extensions).

On a whim I decided to select a server. I used R.I.T. (another local college). Ping times were in the upper 30 to lower 40 msec range BUT speeds were always correct - ~220 upload 11 download.

Here is a graphic - guess when I switched spdMerlin to fixed/R.I.T.

I also tried the fixed server on the Asus built in - same success.

I switched both back to Auto - both selected University or Rochester - and both showed the erratic download speeds.

I went through the Asus speedtest dropdown server list and tried a few more. 3 out of 8 gave the erratic results.


Moral of the story - shorter pings don't always mean the best/fastest server. If your speedtest results start to look erratic, try a few different servers...

 

[kfahoo]

fyi

Choose an option:    2

Generating list of closest servers for WAN...

/jffs/scripts/spdmerlin: line 2990: jq: not found
[: bad number
/jffs/scripts/spdmerlin: line 2990: jq: not found
/jffs/scripts/spdmerlin: line 2990: jq: not found
...

I noticed today, right after i updated my AX86U to the final 386.1 firmware from Merlin that the Add-Ons webUI tab is not there anymore... and i can't access spdMerlin, any suggestion? I tried update and force update, and it did nothing to the webui? Everything else is fine... i am worried if i uninstall and then re-install it might fix it, but i will lose my gathered data, right?

had same problem, I've reset amtm and it's fine now

 

[iTyPsIDg]

The good news, everything seems to be working as expected now that I excluded it from QoS.

I do have another question about AutoBW. Would it be possible to give that a separate schedule from spdMerlin checks? Here is the use case: I would like results of speed tests throughout the day so that I can see if anything weird happened at any point; however, I don't want the QoS values changed while I'm in the middle of a meeting because it resets the firewall and drops my connection.

 

[itibi]

** Ookla vs nPerf vs speedtest give different results **

Hi! Recently, I change my modem for a faster one TC 4400 (max speed=940 Mbps) and I also increase my ISP's package to 400 Mbps download max.

The nice addon spdmerlin (executed at 30 minutes interval since the last 12 days) show me an average speed of 300 Mbps with a maximum around 330 Mbps.
right now, spdmerlin show me a current speed of 330 Mbps but when contacting my ISP, they confirm me that the current speed between my modem and their infrastructure is around 450 Mbps!!

I did some tests:
1. My windows 10 PC connected directly on modem using nPerf (web or native) tests show 451 Mbps
2. The same PC connected on 1 Gbps port on my Asus RT-AC3200 port show 453 Mbps
3. spdmerlin on that same router show 326 Mbps

Is this a limitation on Ookla used by spdmerlin ?
Is it due to my router CPU performance ?

I do not have the knowledge to run nPerf unix version of my router, so I cannot show this result.

Any ideas ?
Thanks

 

[L&LD]

If you're using QoS on the router, those results are expected.

 

[JGrana]

** Ookla vs nPerf vs speedtest give different results **

Hi! Recently, I change my modem for a faster one TC 4400 (max speed=940 Mbps) and I also increase my ISP's package to 400 Mbps download max.

The nice addon spdmerlin (executed at 30 minutes interval since the last 12 days) show me an average speed of 300 Mbps with a maximum around 330 Mbps.
right now, spdmerlin show me a current speed of 330 Mbps but when contacting my ISP, they confirm me that the current speed between my modem and their infrastructure is around 450 Mbps!!

I did some tests:
1. My windows 10 PC connected directly on modem using nPerf (web or native) tests show 451 Mbps
2. The same PC connected on 1 Gbps port on my Asus RT-AC3200 port show 453 Mbps
3. spdmerlin on that same router show 326 Mbps

Is this a limitation on Ookla used by spdmerlin ?
Is it due to my router CPU performance ?

I do not have the knowledge to run nPerf unix version of my router, so I cannot show this result.

Any ideas ?
Thanks

I had issues recently with the Ookla method of selecting a server for the speedtest. I believe it uses the shortest ping time. In my case, it caused much lower results with the server it selected.

Try to select your own. Try a few. If they all show the 300Mbps speed, start to suspect (as @L&LD said) things in the router like Adaptive QoS.

 

[itibi]

@L&LD: Thanks for your fast reply.

No, I am NOT using QOS.
Router has been rebooted before testings.

To resume:

test 1: Windows+speedtest ----------> RT-AC3200 --------------------> Modem -------> ISP shows 447 Mbps
test 2: Windows+web(nperf.com) ----> RT-AC3200 --------------------> Modem -------> ISP shows 453 Mbps
test 3: Windows+web(nperf.com) ---------------------------------------> Modem -------> ISP shows 451 Mbps
test 4: RT-AC3200+spdmerlin --------> Modem -------> ISP shows 344 Mbps

On all previous tests, I select the SAME test server (ebox.ca) which is my ISP

@JGrana: good idea, I test spdmerlin with 5 different servers, here are the 5 results:
339 Mbps, 327 Mbps, 332 Mbps, 323 Mbps and 349 Mbps
so they all show the same result.

If we exclude Adaptative QOS from the equation, any other suspect?

 

[JGrana]

@L&LD: Thanks for your fast reply.

No, I am NOT using QOS.
Router has been rebooted before testings.

To resume:

test 1: Windows+speedtest ----------> RT-AC3200 --------------------> Modem -------> ISP shows 447 Mbps
test 2: Windows+web(nperf.com) ----> RT-AC3200 --------------------> Modem -------> ISP shows 453 Mbps
test 3: Windows+web(nperf.com) ---------------------------------------> Modem -------> ISP shows 451 Mbps
test 4: RT-AC3200+spdmerlin --------> Modem -------> ISP shows 344 Mbps

On all tests, I select the SAME test server (ebox.ca) which is my ISP

Is Windows selecting the same speedtest server as spdmerlin? It looks like ebox.ca uses nperf.com technology/servers. spdMerlin/Asus use Ookla.

 

[itibi]

Is Windows selecting the same speedtest server as spdmerlin? It looks like ebox.ca uses nperf.com technology/servers. spdMerlin/Asus use Ookla.

Yes, I tried the same server on both test

I also test it with the windows Speedtest program from Ookla which shows 428 and 438 Mbps on two different servers

So, Ookla running on a PC connected to router shows 428 Mbps
and Ookla running directly on that same router shows 325 Mbps.

btw: windows speedtest program from Ookla version is 1.13.154

 

[dave14305]

So, Ookla running on a PC connected to router shows 428 Mbps
and Ookla running directly on that same router shows 325 Mbps.

Open a router SSH session in one window and run top -d 1. Watch the % idle on the second line. Start the test from the router GUI or another SSH session. But keep watching the number to the left of idle. Report the highest percent you see during the test.

A speedtest running on the router will always be more constrained than a speedtest just passing through the router.

 

[L&LD]

The other suspects are any and all TrendMicro options you may have available and using on your router.

Even if you tested/tried them once, and then turned the feature off.

Go to the Administration, Privacy page, and Withdraw consent for all. Then, reboot.

You can also check what the Tools/Network section shows for HW acceleration too. Before and after the reboot.

 

[itibi]

Open a router SSH session in one window and run top -d 1. Watch the % idle on the second line. Start the test from the router GUI or another SSH session. But keep watching the number to the left of idle. Report the highest percent you see during the test.

A speedtest running on the router will always be more constrained than a speedtest just passing through the router.

@dave14305:
top -d 1 shows:
1. before running WEBUI spdmerlin on WAN with my preferred server (ebox.ca), idle is between 95% - 100%
2. during spdmerlin execution, idle is between 40% - 80%

@L&LD: I will test it tomorrow and report.

Thanks

 

[itibi]

@L&LD: Finally, I got time to testings, here are the results:

FYI: Adaptative QoS is disable

Before testing, a spdMerlin WAN test with ebox.ca show a download speed of 355 Mbps

Also, note that on Asuswrt-Merlin 384.13_10, HW acceleration is under the tab Lan/SwitchControl

Test 1: change Nat Acceleration from AUTO to DISABLE + reboot;
- Now it displays Nat traffic is processed by CPU
- the spdMerlin test show 375 Mbps, so almost no difference

test2: I discard AiProtection Privacy + reboot
- the spdMerlin test show 447 Mbps

I tried various test combinations : (QoS on/off) (NAT acceleration=AUTO/CPU) (AiProtection=on/off)

the only combination which give me over 400 Mbps is QoS=off, NAT acceleration=CPU and AiProtection=off


My first reaction was to check the pertinence of AiProtection
which within the last 30 months blocks 258 malicious sites and 3 attacks from two ways IPS.

So now, I have to put in the balance AiProtection on one side
compare to a gain of 20% of my download speed on the other side.

Any opinion ?

 

[elorimer]

Watching the icing start in our neighborhood, likely soon to lose power. When I do, and I fail over to a very slow cell connection, is there a way to disable spdMerlin (connmon too) and other things that use bandwidth without uninstalling it? It doesn't need to be automatic; I'd just like to go into the webgui and switch it off for the time being.

 

[L&LD]

Yes, you can. From both the GUI and amtm.