QOS question

[Netduma Fraser]

Thanks for that, when you factory reset previously, did you change anything else other than QoS settings? Have you disabled WMM in Advanced Wireless settings by any chance?

[CRarsenxL]

On 6/11/2021 at 7:05 AM, Netduma Liam said:

OK, in Traffic Prioritisation, could you please add a new rule for your device that you're running ping plotter on and select ICMP-Ping. This will now be a more accurate representation of what you're seeing in game. Could you then run a ping plotter and share the results with this rule enabled?

The spikes you're seeing there make sense as when you're running a connection benchmark, it's filling up the network queue to see how this is handled by QoS.

So then by prioing ICMP ping — you’re essentially emulating the gaming device on the network ; this way u can actual run a “live test” to see how those packets react?

[Netduma Fraser]

23 minutes ago, CRarsenxL said:

So then by prioing ICMP ping — you’re essentially emulating the gaming device on the network ; this way u can actual run a “live test” to see how those packets react?

You're emulating the way they determine your ping but not the actual gaming traffic - it's independent of that

[WACK]

Could it be a bad modem?

 

[WACK]

On 6/11/2021 at 2:15 PM, Netduma Fraser said:

Thanks for that, when you factory reset previously, did you change anything else other than QoS settings? Have you disabled WMM in Advanced Wireless settings by any chance?

I dont think I changed WMM settings, no.

 

[Netduma Fraser]

Could you try downgrading to a pre-DumaOS 3.0 version, factory reset and see if it seems to work better please?

[WACK]

@Netduma Fraser would 2.3.2.114 work? I also see 2.3.2.56 on the forum. I am not sure what is pre3.0

 

[WACK]

So I downgraded to 114

 

It seems I get closer to full speed to my PC https://www.waveform.com/tools/bufferbloat?test-id=7db2b654-fca2-417b-85d5-53241ea46af9
Also, It seems pingplotter isnt getting the massive ping spikes with QOS, but I still cant resolve the issues with pink spikes on DSLreport or waveform

I thought it may help to run some tests at various up and down limits on the QOS settings. I noticed that if upload is at anything but 100, I get a lot of latency:
For example, at 100% on upload, I have a very tight and low latency spread on upload, but if I lower it, it goes through the roof. I would assume if I set it lower, it would help with bloat.

This chart shows my tests at 70% download, and various upload, 95th% ping during upload bloat-test

 

But what is really confusing to me, is that lowering the upload from 100, while leaving download set to 70, has the same effect on the download latency

This chart shows my tests at 70% download, and various upload, 95th% ping during download bloat-test



A little aside, lowering the slider for upload increase median ping for download, and lowers median ping for upload (but, again, at the cost of large ping spikes). 

This seems not-so-bad, if I leave it at 70down/100up for QOS. Is that how its supposed to work? I did a lot of tests and transposed the data into tableau and found that combination produces the best results. Is this the best I can do? I feel like its in a good place via ping plotter and waveform, but a tiny improvement would get me that sexy A+ test. 

[Netduma Liam]

Thanks for this, that's interesting.

Out of interest, what is Connection Benchmark now reporting after the new firmware was installed? Did you also factory reset after the new firmware? If not, this could be a reason we're seeing these anomalies. Ideally you'd like to have the upload slider in Congestion Control lower as currently DumaOS won't have much control.

[WACK]

6 hours ago, Netduma Liam said:

Thanks for this, that's interesting.

Out of interest, what is Connection Benchmark now reporting after the new firmware was installed? Did you also factory reset after the new firmware? If not, this could be a reason we're seeing these anomalies. Ideally you'd like to have the upload slider in Congestion Control lower as currently DumaOS won't have much control.

I dont see connection benchmark on this version. 

[Netduma Fraser]

Tests like Waveform can be negatively impacted by Traffic Prioritization so I'd stick with PingPlotter which is what we normally suggest or disable Traffic Prio completely and retry those tests.

[WACK]

OK, checking back in. If ANYTHING is connected to the router (wemo light switch, ring doorbell - even inactive, wifes phone, my phone, my computer, etc), I get ENORMOUS latency spikes playing games. if I block everything I get really good games. 

I have my ISP coming out again tomorrow, and then I will be switching out my equipment (modem and router). I feel like I am better off with a ISP provided gateway. Is there a reason to buy another router with dumaOS? It sounds like you guys have no idea why QOS, across multiple versions and settings, makes my connection worse. I had my eye on the R2 since I dont really use wifi much, but the XR1000 also seemed like an option. I just cannot believe with all the testing I have done the answer is "retry the tests." I did (actually, with keeping records) 100+ tests with various QOS settings and software versions,  and they all show QOS makes it worse. At best QOS doesnt make it better.  

Is this a router problem? Is it DumaOS not working? Is it my Modem? Is it the ISP? Honestly, I am at the point where gaming just isnt fun anymore. 

I appreciate the help troubleshooting, but this unit seems straight up broken. 

[Netduma Liam]

7 hours ago, WACK said:

OK, checking back in. If ANYTHING is connected to the router (wemo light switch, ring doorbell - even inactive, wifes phone, my phone, my computer, etc), I get ENORMOUS latency spikes playing games. if I block everything I get really good games. 

I have my ISP coming out again tomorrow, and then I will be switching out my equipment (modem and router). I feel like I am better off with a ISP provided gateway. Is there a reason to buy another router with dumaOS? It sounds like you guys have no idea why QOS, across multiple versions and settings, makes my connection worse. I had my eye on the R2 since I dont really use wifi much, but the XR1000 also seemed like an option. I just cannot believe with all the testing I have done the answer is "retry the tests." I did (actually, with keeping records) 100+ tests with various QOS settings and software versions,  and they all show QOS makes it worse. At best QOS doesnt make it better.  

Is this a router problem? Is it DumaOS not working? Is it my Modem? Is it the ISP? Honestly, I am at the point where gaming just isnt fun anymore. 

I appreciate the help troubleshooting, but this unit seems straight up broken. 

My apologies if the troubleshooting process has been frustrating for you, the primary reason for this is because it's not an issue that we've seen before, it doesn't really make sense.

Once you ISP has installed new equipment, let us know how you get on with it and we can proceed from there if you wish to, though I do understand your frustrations.

[WACK]

@Netduma Liam

Are you guys pretty sure this doesnt look like the ISP?

I "asked" the ISP to check the node that is their first hop from my home, and then I got a call "Tier 3" support from them to notify that they now a replacement scheduled and would call back with a date and time this week.

It seems like it would be related?

W

 

[Netduma Fraser]

48 minutes ago, WACK said:

@Netduma Liam

Are you guys pretty sure this doesnt look like the ISP?

I "asked" the ISP to check the node that is their first hop from my home, and then I got a call "Tier 3" support from them to notify that they now a replacement scheduled and would call back with a date and time this week.

It seems like it would be related?

W

If it was that then the issue would occur regardless of the device you were using which is why it doesn't really make sense. Perhaps a weird incompatibility but weirder things have happened, I'd be interested to see if it does make a difference

[WACK]

Hey @Netduma Fraser

 

So I have been talking with ISP about the node and pedestal, and as part of their troubleshooting they upgraded my plan to 1gbps down and whatever lousy speed up.

When they test it, it gets roughly 1gbps, and their open bandwidth test shows my area tops at 1.3gbps. My PS5 shows ~800mbps, I dont trust my PC for speedtesting, but the built-in speedtest for the router tops out at ~520mbps consistently. 

When I factory reset with the new service, it showed 942mbps, so I am pretty sure I am getting there. Is there a setting I am missing on the speedtest that would cause it to max out at 500mbps?

I found this comment from you https://community.netgear.com/t5/Nighthawk-Pro-Gaming-Routers/XR500-dumaos-3-0-speed-test/td-p/2094386
>We're still improving the results reported by Connection Benchmark so you can ignore that for the moment, if a third party speed test gives you the speeds you expect that is what you'll be getting
 

Maybe just a known issue and ignore it?
 

[Netduma Fraser]

@WACK yeah known issue, you can ignore the speed test result on the XR routers for the moment - sometimes it shows close to what you expect sometimes it doesn't - sounds like you are getting expected speeds elsewhere

[Bert]

Your issues are with your ISP not the router.

 

It's the way these systems are managed and there is very little you can do about it.

 

Your modem is connected to their CMTS along with 500ish other customers. This CMTS has about 50mbit per channel available in DOCSIS 3.0 mode. So in theory, 32 downstream channels is 1.6GB/s, this is for all the 500 users. However, depending on their frequency spread, they might have more than 32 downstream channels increasing that capacity a bit.

 

Now, when you fire up a download. 10 other subscribers are downloading as well at 160mbit. Their CMTS needs to make space for your download and then load balance this so you all get a fair share. That's that ping spike you are seeing right at the start of a download. There is nothing your router can do against this. 

 

What makes these cable systems very hard to smooth out is the variable link speed involved. Say you have 600mbit subscribtion. But you might only have 200mbit at that exact time.

 

I have replaced my Puma6 modem with a Broadcom one and I get much of the same as you get. ISP's can program different modems with different settings and they include QoS and traffic shaping as well. My broadcom modem is normally only used for business lines and has no form of traffic shaping or QoS on it. That means that if you do a speedtest and you exceed their maximum bandwith, their traffic policer hard enforces your bandwith giving you a massive ping spike. That is not so much of an issue, the issue is that you don't really know where this limit will be since your amount of available bandwidth depends heavily on the other users on the netwerk. Ie it's impossible to find a set and forget setting as it litterally changes every 5 minutes.

 

Good side is that these Broadcom modems are much more responsive over their intel Puma counterparts, if you don't do any crazy stuff they will work very well.

 

That graph you see in connection benchmark, on upload. That is true bufferbloat. Latency running up while it's filling the buffer on the modem / ISP equiment. DumaOS routers do traffic shaping as well but on connection benchmark it bypasses this and hits the ISP limits. That's why the graph is so crappy.

 

Also the way this works you don't want to use auto setup and use your ISP's overprovisioning. If your subscription is 400 / 10 but you get 470 / 12 in the test, it's better to set it to 400 / 10 since that is what the router bases it's traffic shaping on.

[WACK]

@Bert

Thank you.

You say that there is nothing the ISP can do. The people helping me are responding to an FCC complaint now, and have floated the idea of too much usage on the line in my area. They are insanely responsive now, and have whatever an "executive support" liaison, and a business technician regional supervisor (who brought over test account and his own XR500 to help me troubleshoot. He took the XR500 home, but left the test account modem.) working directly with me. 

Like I said, on a completely uncapped test of the line in the street, the max down is ~1300mbps and the upload in around 110mbps. Pretty narrow upload if people are pushing web traffic from the dozens of homes, I would thing, especially with all the cameras in the area. 

The supervisor also mentioned that the upload being so limited on their system is a major issue internally since the pandemic started that is part of their 4 year plan to rectify. That is also a reason he suggested moving from 400 to the 1.2gbps plan to increase my 10up to ~40up. In other words, it sounds like you and the guy I am talking to are saying similar things. He thinks the low upload is causing ping spikes. 

What do you suggest I ask from them, or how to talk to them? They have an order to replace the pedestal already because they think traffic might be getting jammed in that, and have talked about the neighborhood maybe not having sufficient bandwidth.  I am not a technician, so I am not sure what to tell them.