FlexQoS FlexQoS issues with 388.4 HND5.04 models

[Kingp1n]

Testing QoS effect with the router's built-in speed test tools often generates wrong results for various reasons. You guys test with actual client, wired preferably to eliminate dependent on available bandwidth (active clients) variable Wi-Fi latency. Testing on Wi-Fi - you can play with sites and QoS knobs forever. In most scenarios the ISP applies QoS on their end, you are fighting on your end and Wi-Fi has QoS on it's own on top. The final result - good luck. The final solution - better ISP and fiber when available.

While this may be true with certain type of speeds (500Mbps+ possibly) & certain type of routers (regardless of brand)....yes it's always best to test with wired devices vs testing with wireless clients since it will always provide different speeds results.

I think it's unfair to make this comment on using the built-in speedtest. Also and and as you mentioned....things like your specific ISP plus anything else specific to their environment (is QoS enabled / AiProtect / additional scripts ect...) can play a role to testing Internet speeds (wired vs wireless).

I'm not sure what your paid speeds are but I believe you no longer use Asus routers as your daily driver.

From my own experiences/environment with my internet speeds/current setup and my Asus router, using the internal GUI built-in speedtest have always given me accurate results.

 

[Tech9]

I think it's unfair to make this comment on using the built-in speedtest.

You have to be aware how the selected QoS manipulates the traffic in both directions.

you no longer use Asus routers as your daily driver

I never used any Asus routers as daily drivers. This allows me to explore different behavior on test only units without disturbing my network.

 

[Kingp1n]

I never used any Asus routers as daily drivers. This allows me to explore different behavior on test only units without disturbing my network.

I used plenty of different brands in the past as daily drivers. This allows me also to explore different behaviors on real life units with or without disturbing my network. Only way to truly find out!

 

[Tech9]

Only way to truly find out!

We obviously have different ways to find out what works and what doesn't. This is okay for me. Seems like you didn't catch what my speed test message was about. It wasn't about commonly known CPU limitations and what else is running there on the CPU. It was about how the selected QoS method manipulates the traffic and the effect on built-in speed test in particular. If you believe this doesn't apply to your particular setup - this is fine. Testing with a wired client is the more accurate method though.

 

[Kingp1n]

We obviously have different ways to find out what works and what doesn't. This is okay for me. Seems like you didn't catch what my speed test message was about. It wasn't about commonly known CPU limitations and what else is running there on the CPU. It was about how the selected QoS method manipulates the traffic and the effect on built-in speed test in particular. If you believe this doesn't apply to your particular setup - this is fine. Testing with a wired client is the more accurate method though.

I did say testing results will vary whether wired or wireless. What works for me doesn't mean it will work for others.

Not sure there's anything else/more to add. Each environment will vary. Yes testing with a wired client is probably more accurate. I did say from my own experiences/environment with my internet speeds/current setup and my Asus router, using the internal GUI built-in speedtest have always given me accurate results.

 

[Tech9]

Each environment will vary.

The reason I prefer test methods excluding or at least minimizing the environment factor.

If something fails consistently in tests there is no way it miraculously fixes itself and starts working in "real life" application.

 

[Kingp1n]

I get the feeling you don't like Asus products(routers) and that's fine too.

I can only speak from my own experiences and for me Asus routers have fixed alot my issues in my "environment" when compared to other brands.

I don't have the time to keep "testing" routers" to see if they will fail. I use them real-life and once they fail I move to another. I don't stay loyal to any brand. That's okay too.

 

[Tech9]

I get the feeling you don't like Asus products

I won't comment feelings. Unrelated to the conversation. The question:

What is the better test method for QoS - built-in speed test or wired client?

Straight answer - wired client. Asus router or other vendor - the same answer.

I don't have the time to keep "testing" routers

I know. You find what works and what doesn't on the fly like most folks around.

 

[Kingp1n]

I won't comment feelings. Unrelated to the conversation. The question:

What is better test method for QoS - built in speed test or wired client?

The answer - wired client. Asus router or other vendor - the same answer.

You remind me of another member (@sfx2000) that seem to know everything about Asus routers. I simply stated my own experience and that doesn't mean I'm correct/works for me will work for all. I don't think I mentioned anything out of the ordinary.

I'll let you have the last comment as it seems you know it all when it comes to Asus routers even though you don't even use one as a daily driver.

 

[Tech9]

Your post has more "you" than the subject we discuss - "QoS". Not needed.

Asus didn't invent QoS, by the way. The QoS options you see in Asuswrt are not unique to Asus products. As I said - I don't rely on "seems like" and "have a feeling". We are talking about hardware and software and how it works together. There is a reason I don't use Asus or other home AIO routers as daily driver. There are more capable products out there for my needs and expectations. You made your choice, I made mine. Me using time to test your choice is your advantage. The results - take it or leave it.

End of conversation.

 

[Kingp1n]

Your post has more "you" than the subject we discuss - "QoS". Not needed.

Asus didn't invent QoS, by the way. The QoS options you see in Asuswrt are not unique to Asus products. As I said - I don't rely on "seems like" and "have a feeling". We are talking about hardware and software and how it works together. There is a reason I don't use Asus or other home AIO routers as daily driver. There are more capable products out there for my needs and expectations. You made your choice, I made mine. Me using time to test your choice is your advantage. The results - take it or leave it.

End of conversation.

As I said... "you” had the last comment. I was done with this conversation

 

[sfx2000]

You remind me of another member (@sfx2000) that seem to know everything about Asus routers. I simply stated my own experience and that doesn't mean I'm correct/works for me will work for all. I don't think I mentioned anything out of the ordinary.

Leave me out of this - I'm an outside observer/developer that chimes in from time to time...

 

[sfx2000]

I think it's unfair to make this comment on using the built-in speedtest. Also and and as you mentioned....things like your specific ISP plus anything else specific to their environment (is QoS enabled / AiProtect / additional scripts ect...) can play a role to testing Internet speeds (wired vs wireless).

It's not just me that considers "speed testing" on the router itself is not a good representation of performance overall...

It is always best to do it from the client side, as this is real-world traffic, whether it's over the wire or wireless...

Cloudflare's test is pretty good, as is Waveform - speedtest.net doesn't push enough traffic over enough time to get a good feel...

Even something as simple as iperf3 can get misleading results.

 

[BikeHelmet]

I only have AX routers now. I never had to disable anything to get QOS working on older ones.


Probably Other. Something above the slowest classes, in case things get mis-prioritized or take a little bit to get categorized.


What does it show on the QOS page as your limit? What are the category limit %'s in FlexQOS?

It seems probable that if you're getting a low speedtest result, consistently at a low level, it may be getting mis-categorized and throttled. We need to see the full QOS configuration to tell you for sure. It might be worthwhile to grab a screenshotting tool like GoFullPage and upload some screenshots. It scrolls, screenshots and stitches into a single PNG.

 

[Kingp1n]

It wasn't about commonly known CPU limitations and what else is running there on the CPU.

Speaking of CPU limitations... I've only owned the "Pro" version for a few months. I haven't ran too many add-on scripts (see signature) but I noticed the RAM was running at around 93-94% (nothing unusual since this is normal) after a few days. Router running smoothly. I decided to run a speedtest today thru the internal GUI.

Suddenly, my GUI froze and the WiFi restarted (signal dropped). However the router did not restart. I did have to log back in to GUI and the speedtest did not complete. I decided to remove Flex, turned off QoS & I ran another speedtest...same issue (GUI froze/WiFi signal dropped). It's definitely not with QoS enabled and/or running the Flex script (running at default settings). A router restart fixes the issue as the RAM goes back down to around 60%.

This weekend, I'll troubleshoot whether the issue maybe Diversion (also default settings) running with Guest Network/ AiProtect/QoS enabled... possibly limiting the CPU power/choking the speedtest forcing the WiFi to restart ONLY after running an internal speedtest thru the GUI! It seems it only happens when RAM is the 90%ish (above)!

I never had this issue with the GT-AX11000 "non-pro" version. It seems they dont make the new hardware as good as the original! Initially I thought I got a faulty router thru Amazon but the replacement is doing the same thing! Moral of story don't use internal speedtest at least until I can troubleshoot was causing it.

 

[Tech9]

I don't have Pro model to test with. I know people usually respond with clichés like here in this thread, but I also know Asus routers above 90% RAM utilization may run into issues. Linux may have good RAM management, but the pyramid of accumulated over many years code called Asuswrt wins. I can crash most Asus routers with specific file transfer without touching them. They run out of memory and start shutting down essential services. Some recover, some don't and require reboot. The blunt honest answer - I don't care. The community is overly sensitive about their toys and... I don't use Asus routers on my networks.

 

[BikeHelmet]

I don't have Pro model to test with. I know people usually respond with clichés like here in this thread, but I also know Asus routers above 90% RAM utilization may run into issues. Linux may have good RAM management, but the pyramid of accumulated over many years code called Asuswrt wins. I can crash most Asus routers with specific file transfer without touching them. They run out of memory and start shutting down essential services. Some recover, some don't and require reboot. The blunt honest answer - I don't care. The community is overly sensitive about their toys and... I don't use Asus routers on my networks.

My experience has been that there are dealbreaker firmware versions. I'm always very careful rolling out a new firmware release. I do it on my home network and evaluate for a few months before doing it in the wild. Entware can also be problematic - I don't always update that. That has spared me a lot of grief. Asuswrt has quite a few gotchas, especially once modded, but the routers can be ludicrously stable as well due to their Tomato origins.

Last week I got an office building up and running again, same day service, but had to use what I had onhand. Their pfSense router melted after 2 years of ownership. Cooked its brains and decided that 10mbit was a proper speed for switching. Login page? "What's a login page?" I ended up swapping in an RT-AX86U. One of their other buildings has the same router and has been working flawlessly for 2.5 years.


@Kingp1n Start with AI Protect. It's glitchy and can take down the whole router. If you turn that off, the memory leak might go away.

 

[Tech9]

One of their other buildings has the same router and has been working flawlessly for 2.5 years.

Did they catch the broken firewall in Asuswrt for >1 month or not?

Home router running "office building" network sound like Space X buying a part from Home Depot when they don't have the right one.

 

[BikeHelmet]

Did they catch the broken firewall in Asuswrt for >1 month or not?

Home router running "office building" network sound like Space X buying a part from Home Depot when they don't have the right one.

These aren't huge office buildings, but are using /22's, so they're not puny either. Interesting examples that you chose there. SpaceX is well known for using consumer parts where they can. They designed their own motherboards for the Falcon 9 rockets that used three standard dual-core Intel CPUs... if they can source a commodity valve or pipe, they do. RocketLab's whole launch/mission rooms are chock full of regular electronics that you can buy off Amazon. No fancy custom designed stuff there - just off the shelf stuff plugged into sensor boards. Probably networked Raspberry Pis. They're trying to make commodity parts satellites their business model, so it makes sense.

It's pretty common practice, because you as the ultimate deployer can better use-case-test than the manufacturer that has many different customers utilizing something for different business cases. They are less likely to properly predict what to test for to avoid failure than the actual person deploying is.

I have a lot of experience with Tomato routers going right back to the pre-Toastman days, so have pretty good knowledge of what to avoid in a business use case. There is no QOS and no AI Protection and no on-router services like VPNs. Nothing that would memory leak. Ports/security are cranked up a bit. It's really just a well heatsinked well accelerated routing machine. Where capabilities are lacking, other devices fill in - like network switches that intercept rogue routers/DHCP and WAPs. These office spaces have a lot of BYOD, so even though most devices have some sort of endpoint security, you must still treat the whole network as if you're in the public and attacks could come from anywhere.

 

[Kingp1n]

@Kingp1n Start with AI Protect. It's glitchy and can take down the whole router. If you turn that off, the memory leak might go away.

Thanks @BikeHelmet. No changes on my setup yet but I decided to install RTRMON to keep an eye on the health of router status. Currently RAM is at 94% and the speedtest thru RTRMON successfully completed. I still have not done one thru internal GUI. I'll update if I decide to do so. Router is running smoothly & no signs of memory leak at this time! Running uptime = 5 days!!!