spdMerlin spdMerlin - v4.x - speedtest and bandwidth monitoring

[s_Fanous]

So I just got a 1.5Gbps connection and installed spdMerlin to test it. I don't have any 10Gbps capable devices at home and am using an AC-RT68U so understand that the router itself can only handle 1Gbps.

spdMerlin installed is v4.3.0. I don't have any QoS setup.

spdMerlin is setup to run once every hour and is consistently reporting around 250Mbps only. Now if I execute the test from a browser from a device directly connected to the router over RJ45 I get close to the 1Gbps limit. I've also tested using WiFi and the Speedtest app on iOS and get around 600Mbps.

Why is spdMerlin only managing 250Mbps while other tests are reporting much higher speeds?

 

[L&LD]

What 1.5Gbps ISP connection did you get? Some only promise that speed to two devices or more. A single device can't ever hit that speed (even if it has a 2.5GbE, 5GbE, or 10GbE connection).

You don't mention your router model, but all models are limited when running greater than 1Gbps speeds, internally. Simple solution; just ignore those results.

 

[Lynx]

What 1.5Gbps ISP connection did you get? Some only promise that speed to two devices or more. A single device can't ever hit that speed (even if it has a 2.5GbE, 5GbE, or 10GbE connection).

You don't mention your router model, but all models are limited when running greater than 1Gbps speeds, internally. Simple solution; just ignore those results.

Agreed (all these people with 1Gbps connections moaning about speeds when many others like me make do with sub 100Mbps on LTE!), but am curious to understand why the lower result in spdMerlin. Mine are all completely consistent.

 

[s_Fanous]

What 1.5Gbps ISP connection did you get? Some only promise that speed to two devices or more. A single device can't ever hit that speed (even if it has a 2.5GbE, 5GbE, or 10GbE connection).

You don't mention your router model, but all models are limited when running greater than 1Gbps speeds, internally. Simple solution; just ignore those results.

It's Bell Fibe 1.5Gbps in Canada. To be honest I'm not too fussed with evening hitting these speeds on a single device. But it still doesn't explain why spdMerlin reports 250Mbps, whereas from a device connected to the router it's in the 800-900Mbps.

My ISP supplied router is a Bell Home Hub 4000, whereas my Asus Router is the RT-AC68U running Merlin V384.19. I directly connect the Asus Router to the ISP using PPPoE.

BTW, if I speed test from the Bell HH4000 itself it reports 1.5Gbps.

 

[L&LD]

Of course, ISP-supplied equipment will report full speed.

I don't think spdMerlin supports 384.19.

Even if it does, RT-AC68U is woefully underpowered to hint at 1.5Gbps speeds.

 

[s_Fanous]

Of course, ISP-supplied equipment will report full speed.

I don't think spdMerlin supports 384.19.

Even if it does, RT-AC68U is woefully underpowered to hint at 1.5Gbps speeds.

I know ISP equipment will always be biased, but that still doesn't explain the discrepancy between spdMerlin on the router and Speedtest from a device connected to the Asus router.

Even a real world test proves that RT-AC68U can handle more than 250Mbps. I just downloaded the Ubuntu image using BitTorrent and hit over 70MBps (So about 560Mbps)

I didn't know spdMerlin didn't support 384.19. Must have missed that. That might be the culprit then.

 

[L&LD]

Running a speedtest on the router itself and using the router to run a speedtest on a connected device is not the same thing.

 

[Jack Yaz]

Running a speedtest on the router itself and using the router to run a speedtest on a connected device is not the same thing.

This. The router CPU is weaker than that of modern mobile phones. The 68U is quite limited in CPU so asking it to run a speedtest is a lot. Clients of the network benefit from hardware acceleration (read: hacks) that allow traffic to bypass the cpu and give you the gigabit speeds. If you enable qos or disable hardware acceleration speeds decrease because the cpu now has to process every packet and client speeds drop

 

[s_Fanous]

This. The router CPU is weaker than that of modern mobile phones. The 68U is quite limited in CPU so asking it to run a speedtest is a lot. Clients of the network benefit from hardware acceleration (read: hacks) that allow traffic to bypass the cpu and give you the gigabit speeds. If you enable qos or disable hardware acceleration speeds decrease because the cpu now has to process every packet and client speeds drop

Thanks. That makes sense now.

 

[StroppyOz]

Does QoS on vs off change the bandwidth reported by the browser?

Less then spdMerlin QoS off but relatively consistent. Here's the data:

QoS Off:

• Router
• Browser

QoS On:

• Router
• Browser

I have also confirmed that QoS works in the browser (but not router) by shaping upstream/downstream to 50/50.

 

[StroppyOz]

Should I take the absence of any replies to indicate the issue I'm having no-one else has? i.e. it's a local setup issue?

 

[Jack Yaz]

Should I take the absence of any replies to indicate the issue I'm having no-one else has? i.e. it's a local setup issue?

sorry, i've been a bit busy. with adaptive qos, spdmerlin attempts to bypass it by adding a specific mark to the packets so that it goes into the "highest priority" container. unfortunately, packet tagging as used by adaptive qos means the CPU has to process it all, so will still slow speeds down, particularly when the traffic originates on the router. clients don't see as much of a hit as asus/trendmicro have some black magic that lets hardware acceleration co-exist to some degree with qos - router originating traffic can't use the hardware acceleration

 

[Paliv]

What 1.5Gbps ISP connection did you get? Some only promise that speed to two devices or more. A single device can't ever hit that speed (even if it has a 2.5GbE, 5GbE, or 10GbE connection).

You don't mention your router model, but all models are limited when running greater than 1Gbps speeds, internally. Simple solution; just ignore those results.

With Comcast’s gateway I can hit 1.45 Gbps easy on my PC with 2.5GbE. Or did you mean with the ASUS router? Alas, they don’t let you control anything yourself on the router side, so I decided to go back to Merlin.

 

[Jack Yaz]

v4.4.0 is now available
Changelog:

• NEW: On routers that support Asus' built-in speedtest, spdMerlin will now use the same binary
• CHANGED: service-event hook is more selective when it calls spdMerlin
• CHANGED: Licence acceptance is prompted on script install and must be accepted for spdMerlin to be installed.

 

[thecheapseats]

thanks for the alt-location of the License acceptance, Jack...
no ax86u problems now...

 

[Stephen Harrington]

NEW: On routers that support Asus' built-in speedtest, spdMerlin will now use the same binary

@Jack Yaz

I'm seeing some "cosmetic" weirdness since this upgrade.
When I used to run a "manual" speed test from the GUI, I would see an animated progress bar for each interface that was being tested.
With v4.4.0 I get the "Speedtest in progress ..." message and then a blank box below that whilst test is in progress and then the results all sort of spit out at the end fully formed, if that makes sense. Have tried emptying browser cache but same result.
Is this related to the change in speedtest binary or Some Other Magic ??

 

[Jack Yaz]

@Jack Yaz

I'm seeing some "cosmetic" weirdness since this upgrade.
When I used to run a "manual" speed test from the GUI, I would see an animated progress bar for each interface that was being tested.
With v4.4.0 I get the "Speedtest in progress ..." message and then a blank box below that whilst test is in progress and then the results all sort of spit out at the end fully formed, if that makes sense. Have tried emptying browser cache but same result.
Is this related to the change in speedtest binary or Some Other Magic ??

Most likely. Do you see the progress bar on the command line?

edit: found the bug, working on a fix (WebUI bug)

 

[Jack Yaz]

v4.4.1 is now available
Changelog:

• FIXED: WebUI progress for speedtests using built-in binary were missing

 

[Stephen Harrington]

FIXED: WebUI progress for speedtests using built-in binary were missing

All good on 4.4.1 @Jack Yaz, thanks!

 

[StroppyOz]

sorry, i've been a bit busy. with adaptive qos, spdmerlin attempts to bypass it by adding a specific mark to the packets so that it goes into the "highest priority" container. unfortunately, packet tagging as used by adaptive qos means the CPU has to process it all, so will still slow speeds down, particularly when the traffic originates on the router. clients don't see as much of a hit as asus/trendmicro have some black magic that lets hardware acceleration co-exist to some degree with qos - router originating traffic can't use the hardware acceleration

Now I feel like a fool! Totally understand what it's like to be busy.... (he says after replying several days later himself)

Thanks for the explanation. Totally makes sense!