CakeQOS CakeQOS-Merlin

[askan7]

If I leave ping -t while I game it's the same result as meter.com test. Seems like many small packets, like playing apex legends while shooting bots with fast automatic guns causes cake to buffer. No load on network at all, only 1 device and no background services.

Ping to first hop and router while moving and shooting around on Apex Legends.
Cake on:


Cake off:


Can anyone reproduce this? meter.com/ping-test causes the same buffer as certain games as it sends a lot of small packets.
Cake still a lot better with load....

Anyone as any idea on how to fix this?
Can this be caused by my isp's router? It's in bridge mode, but it still gets a public IP and has VLAN for iptv.
I have nothing connected to it besides the asus router but I can't stop it from getting public IP(different from asus public IP) and VLAN.
I might consider to get a cheap dsl modem to see if I get less bufferbloat

 

[heysoundude]

is your router signing in to your ISP or is the gateway/"modem"? your router should be doing all of it if the "modem" is truly bridged - a bridge simply converts signals from one format to another...DSL to ethernet, for instance...and your router can step up to do the connection authentication with your ISP rather than the modem
If the modem is getting a WAN IP, and then assigning an IP to your router to act as its WAN IP, you're double NAT - once at the modem to the router and then again between device and router.
Network Address Translation takes time and that time is latency...in each direction...twice...

Call your ISP's tech support line, or look in their online documentation, for the method to bridge your modem into a true gateway, then set up authentication on your router. Once you've got that handled, then you can tweak cake. (this is a good time to implement Native IPv6 if your ISP offers it...and Merlin can also do IPTV VLAN stuff...)

 

[Morris]

Anyone as any idea on how to fix this?
Can this be caused by my isp's router? It's in bridge mode, but it still gets a public IP and has VLAN for iptv.
I have nothing connected to it besides the asus router but I can't stop it from getting public IP(different from asus public IP) and VLAN.
I might consider to get a cheap dsl modem to see if I get less bufferbloat

Instructions here:

CakeQOS - CakeQOS-Merlin

Interesting unintended consequences for people (ab)using betterspeedtest.sh. https://lists.bufferbloat.net/pipermail/bloat/2020-October/015941.html LOL - how many of the 470-odd blacklisted IPs could be tied to people from these parts would you reckon? (I have no shame using @Jack Yaz 's...

www.snbforums.com

Good luck,

Morris

 

[askan7]

Instructions here:

CakeQOS - CakeQOS-Merlin

Interesting unintended consequences for people (ab)using betterspeedtest.sh. https://lists.bufferbloat.net/pipermail/bloat/2020-October/015941.html LOL - how many of the 470-odd blacklisted IPs could be tied to people from these parts would you reckon? (I have no shame using @Jack Yaz 's...

www.snbforums.com

Good luck,

Morris

No matter the settings, cake always gives ping spikes while gaming with no load. Asus adaptive qos with fq-codel was no such issues but I have worse bufferbloat while connection is loaded.

I also tried everything to fix the random ping spike when a high load starts, like speedtest. Even limiting bandwidth to half doesn't fix it. With cake. fq-codel, doesn't matter. I'm starting to think the problem is the modem/router the ac86u is connected to.

 

[Morris]

No matter the settings, cake always gives ping spikes while gaming with no load. Asus adaptive qos with fq-codel was no such issues but I have worse bufferbloat while connection is loaded.

I also tried everything to fix the random ping spike when a high load starts, like speedtest. Even limiting bandwidth to half doesn't fix it. With cake. fq-codel, doesn't matter. I'm starting to think the problem is the modem/router the ac86u is connected to.

What you describe sounds like link and/or router overload on your ISP's network. Run some traceroutes to various reliable hosts and you will probably see it.

 

[askan7]

What you describe sounds like link and/or router overload on your ISP's network. Run some traceroutes to various reliable hosts and you will probably see it.

You might be right... I did just that while running speedtest with fq-codel:




Depending on the speedtest I use it can spike up to 150ms while the first hops only spike up to 30-40ms.

 

[Morris]

You might be right... I did just that while running speedtest with fq-codel:




Depending on the speedtest I use it can spike up to 150ms while the first hops only spike up to 30-40ms.

You have 30ms of jitter to the first hop. Is it like that without a load?

 

[askan7]

You have 30ms of jitter to the first hop. Is it like that without a load?

Only with load. Without load jitter only happens with cake and I need to run something like a game that sends a lot of small packets

 

[Morris]

Only with load. Without load jitter only happens with cake and I need to run something like a game that sends a lot of small packets

Have you corrected the conflicting queuing methods? only use besteffort

 

[askan7]

Have you corrected the conflicting queuing methods? only use besteffort

I only used besteffort + bridged-llcsnap and of course the limits. Tried a lot of other different things, same issue with idle latency while gaming with cake on.

 

[Morris]

I only used besteffort + bridged-llcsnap and of course the limits. Tried a lot of other different things, same issue with idle latency while gaming with cake on.

Try your pings from the router or another wired computer other than the one you are gaming on. Also, is the computer you are gaming on wired?

Morris

 

[askan7]

Try your pings from the router or another wired computer other than the one you are gaming on. Also, is the computer you are gaming on wired?

Morris

Gaming computer is actually wireless(motherboard built-in intel card), I know... But wireless 5ghz seems pretty stable to me, I have 15m flat cable but it's cleaner this way.

Ping to 1.1.1.1 and wired local pc using cake while gaming. Keep in mind i tried with ethernet and same idle ping spikes happen. Not pings to router/lan, that's intel wifi problem, but they rare and small.


Now with fq_codel and for longer:

 

[Morris]

I think you are trying to fix what is not broken. I see that .3 sec ping in the middle of what is fine yet .3 is not bad particular as it's only happen extremely rarely though it dose happen occasionally. The reason you even notice that 3 ping is it is working much better than it would with no QOS.

Morris

 

[Audionut]

I think you are trying to fix what is not broken.

Agreed.

I would be questioning how noticeable the spikes are to actual game play. Try not to look to hard at it. Don't try and find a problem with the ping spikes with cake while gaming. Just game away and go by your gut feeling afterwards.

 

[askan7]

I think you are trying to fix what is not broken. I see that .3 sec ping in the middle of what is fine yet .3 is not bad particular as it's only happen extremely rarely though it dose happen occasionally. The reason you even notice that 3 ping is it is working much better than it would with no QOS.

Morris

For me it's broken since this is gaming with no load at all. Cake shouldn't increase idle latency at all over no QOS. Look at average ping is like 2ms higher only due to jitter, look at graph, SOO MUCH jitter with cake. With fq-codel or no qos at all, ping/jitter is the same with no load while gaming.

Cake on, 10min test similar to fq_codel test. Previous cake test had different ping interval. Apex Legends is running at stable 144fps in both tests.



fq-codel:



cake:



If the game is closed jitter is the same, leading me to believe cake can't handle small packets (games) well. Or maybe it's because NAT has to run on the CPU with cake. HW acceleration off.
It would be nice if @ttgapers could give some insight on what's going on.

 

[Kingp1n]

For me it's broken since this is gaming with no load at all. Cake shouldn't increase idle latency at all over no QOS. Look at average ping is like 2ms higher only due to jitter, look at graph, SOO MUCH jitter with cake. With fq-codel or no qos at all, ping/jitter is the same with no load while gaming.

Cake on, 10min test similar to fq_codel test. Previous cake test had different ping interval. Apex Legends is running at stable 144fps in both tests.



fq-codel:



cake:



If the game is closed jitter is the same, leading me to believe cake can't handle small packets (games) well. Or maybe it's because NAT has to run on the CPU with cake. HW acceleration off.
It would be nice if @ttgapers could give some insight on what's going on.

Can you post the settings you're using for cake? Maybe you already did and I missed it....
[1] --> Download Speed | [xx Mbit]
[2] --> Upload Speed | [xx Mbit]
[3] --> Queue Priority | [diffserv4]
[4] --> Extra Download Options | xxxxx]
[5] --> Extra Upload Options | [xxxxx]

 

[askan7]

Can you post the settings you're using for cake? Maybe you already did and I missed it....
[1] --> Download Speed | [xx Mbit]
[2] --> Upload Speed | [xx Mbit]
[3] --> Queue Priority | [diffserv4]
[4] --> Extra Download Options | xxxxx]
[5] --> Extra Upload Options | [xxxxx]

You can see it on the printscreen.
[1]. 11.0 down
[2] 0.8 up
[3] besteffort
[4] bridged-llcsnap
[5] bridged-llcsnap

No downloads or uploads going, fq-codel or no QOS at all gives me the same idle latency while gaming.
This only happens if I enable cake.

 

[Audionut]

I think you should be more concerned with the level of packet loss and the inherit problems associated with.

 

[askan7]

I think you should be more concerned with the level of packet loss and the inherit problems associated with.

There's 0 packet loss. First hops packet loss only means they don't respond to the pings, because i'm sending too many of them 10 per second. Everything gets to final hop, and that's what matters. If i set ping interval to 1second everything responds to all pings.

The Difference Between Latency & Packet Loss

What are the differences between latency and packet loss? In this article, we also cover patterns to look for that indicate poor performance.

www.pingplotter.com

Any packet loss signals a problem
Internet protocols are designed to correct some amount of packet loss, but anything over 2% usually effects network performance in a negative way.

When interpreting network performance data, keep in mind the final hop must be effected for anything else to matter. Anything you see that does not carry through to the final hop in the route was sent as a distraction by the demons of bad networking mamba jamba.

 

[Audionut]

If i set ping interval to 1second everything responds to all pings.

Fair enough, I guess.

My asus router has no issues with the frequency of ping signals sent, and I would still consider it a matter that should be investigated. I'm also wary of those who use terms such as "bad networking mamba jamba", but concede the author might simply not want to get into great detail about the subject.

How do I interpret packet loss identified by WinMTR

I've been asked to look into a packet loss problem a previous consultant identified at a client. They have two ADSLMax connections, one for data and the other dedicated to VoIP. The packet loss is

serverfault.com

Basically, he is right on the properties of dropped ICMP packets, although it would be rather untypical to see dropouts on a local Linux machine - especially in the range of 10%.

And I'm still not convinced that the "issue" is an actual issue outside of "omg, the graphs look different", but I will concede that fq_codel does appear to perform slightly better on your network.