Beta Asuswrt-Merlin 386.2 Beta is now available

[Sky1111]

ASIDE: is there a way to get correct speed reading for the devices that are connected to the node?
I either see them as connected via ethernet, or just wrong speed (like is shows AX200 equipped laptop that is actually connected at 2402Mbps, as connected at only 300)

 

[askan7]

@RMerlin
I've done a few more tests with cake while gaming and It's pretty obvious cake is having issues when something that send a lot of small packets is running on the network. Like an online multiplayer game, or a meter.net/ping-test.

Funny thing is that these ping spikes don't show on any in-game ping stats that I've tested (PUBG and Apex Legends) and I don't notice anything while playing either. Soo maybe cake is doing something to ICMP packets? It's not related to having HW acceleration off either as it doesn't happen with Tradicional Qos or Adaptive QOS.

Have a look at this test clip, I've tried to make it as short as possible. I have both the router and pingplotter pinging my first hop while I run an online multiplayer game. Cake on vs off

 

[iTyPsIDg]

A simple example of customizing Cake on Merlin is to use the /jffs/configs/cake-qos.conf.add file. If I want to add the ack-filter option to my upload options, but I don't want to replace the existing firmware-generated options, I can use this syntax:

ULOPTIONS="$ULOPTIONS ack-filter"

This sets the predefined ULOPTIONS variable to itself, then tacks on my ack-filter option. The double-quotes are needed (instead of single-quotes) so that the $ULOPTIONS variable is expanded.

If I wanted to completely replace the ULOPTIONS generated by the firmware, I would use this syntax without including the variable name within the double-quotes:

ULOPTIONS="nat triple-isolate wash ack-filter"

The original variable contents are ignored and replaced with my updated variable when qos is started.

# tc qdisc | grep cake
qdisc cake 8006: dev eth0 root refcnt 2 bandwidth 24576Kbit diffserv3 dual-srchost nat nowash ack-filter split-gso rtt 100ms noatm overhead 18 mpu 64
qdisc cake 8007: dev ifb4eth0 root refcnt 2 bandwidth 215040Kbit besteffort dual-dsthost nat wash ingress no-ack-filter split-gso rtt 100ms noatm overhead 18 mpu 64

I've also added iptables rules for Teams and Zoom to ensure that it takes advantage of the Voice tin in diffserv3 on upload.
This is the contents of my /jffs/scripts/qos-start script. I have IPv6 enabled, so the last 4 lines aren't needed if you do not have IPv6 enabled.

#!/bin/sh

iptables -t mangle -D POSTROUTING -o eth0 -p udp -m multiport --dports 3478:3481 -j DSCP --set-dscp-class EF 2>/dev/null
iptables -t mangle -A POSTROUTING -o eth0 -p udp -m multiport --dports 3478:3481 -j DSCP --set-dscp-class EF
iptables -t mangle -D POSTROUTING -o eth0 -p udp -m multiport --dports 8001:8010 -j DSCP --set-dscp-class EF 2>/dev/null
iptables -t mangle -A POSTROUTING -o eth0 -p udp -m multiport --dports 8001:8010 -j DSCP --set-dscp-class EF

ip6tables -t mangle -D POSTROUTING -o eth0 -p udp -m multiport --dports 3478:3481 -j DSCP --set-dscp-class EF 2>/dev/null
ip6tables -t mangle -A POSTROUTING -o eth0 -p udp -m multiport --dports 3478:3481 -j DSCP --set-dscp-class EF
ip6tables -t mangle -D POSTROUTING -o eth0 -p udp -m multiport --dports 8001:8010 -j DSCP --set-dscp-class EF 2>/dev/null
ip6tables -t mangle -A POSTROUTING -o eth0 -p udp -m multiport --dports 8001:8010 -j DSCP --set-dscp-class EF

Thank you.

I implemented what you have here for my environment. I'm trying to learn more about DSCP for QOS and I found this. Do you think we would be able to use some, or maybe even most, of the recommendations here?

 

[ComputerSteve]

Hi,

Is diversion working with this version?

For me Diversion doesn't work correctly if I enable the Alternate blocking list. If I leave it Standard without fs enabled it works fine.

 

[KingBravery]

@RMerlin
I've done a few more tests with cake while gaming and It's pretty obvious cake is having issues when something that send a lot of small packets is running on the network. Like an online multiplayer game, or a meter.net/ping-test.

Funny thing is that these ping spikes don't show on any in-game ping stats that I've tested (PUBG and Apex Legends) and I don't notice anything while playing either. Soo maybe cake is doing something to ICMP packets? It's not related to having HW acceleration off either as it doesn't happen with Tradicional Qos or Adaptive QOS.

Have a look at this test clip, I've tried to make it as short as possible. I have both the router and pingplotter pinging my first hop while I run an online multiplayer game. Cake on vs off

What ip (10.218.0.1) is that? I can't test it.

 

[dave14305]

Thank you.

I implemented what you have here for my environment. I'm trying to learn more about DSCP for QOS and I found this. Do you think we would be able to use some, or maybe even most, of the recommendations here?

It's much more likely to work on upload versus download.

There was some experimentation before the alpha with implementing Cake on the br0 LAN interface instead of the ifb4eth0 interface, which would allow easier iptables manipulation on download traffic. But performance was dubious.

It may be worth experimenting for those who know what they're doing and how to change it without breaking anything.

 

[askan7]

What ip (10.218.0.1) is that? I can't test it.

That's normal, first hop of my connection outside of the router. It's an internal IP from my isp. It really doesn't matter, even if i ping 1.1.1.1 or 8.8.8.8 it's the same. If there's something like a multiplayer game running i get ping spikes with cake on. Again this doesn't seem to affect the game itself.
It's just weird...

 

[Authority]

Yes it is. Traffic spike.

I re-read this and wondered if you looks at it too fast because your answer didn't make sense. This is not measuring connect speed, this is monthly totals. Is the that you meant?

 

[RMerlin]

I am completely lost.

Set it manually instead of relying on the timezones. It's possible that the router's TZ database isn't up to date, with all the changes that happened these past few years to DST periods.

Really? It's hard to believe they've had a known customer facing issue for years. It doesn't exactly build consumer confidence. One would think if it doesn't work they'd just disable/remove it until it does. Super weird.

It's not as visible on stock firmware because they do not save the database to disk, it gets wiped every reboot.

 

[FTC]

Hi, just upgraded to 386.2 beta1 my mesh system (AX88U router + AX56U node) from 386.1_2 and all is fine except for wifi download speed for 5Ghz AX clients, which seems like capped to <114Mbps or so. I disabled AX mode and wifi download speed returned to 400-500Mbps which is what is to be expected in my network with Aiprotection enabled.

Regarding the system time bug, I could return to GMT+1 time and there was no longer the 1h DST caused displacement bug, however, I can still see the message stating that "the system time zone is different from my locale setting", so I undestand that the fix may not be completed by Asus at this time.

Also while playing with wifi parameters to determine the above commented AX mode problem I saw some entries in the log file of the form:
"kernel: FPM Pool 1: invalid token 0x21588000 freed", which in the past were not a good thing. Those entries seem like not appearing in regular runtime however.

Thanks again to Merlin & team for their hard work!

Will post back more if I see anything else strange.

 

[RMerlin]

I've done a few more tests with cake while gaming and It's pretty obvious cake is having issues when something that send a lot of small packets is running on the network. Like an online multiplayer game, or a meter.net/ping-test.

Try tweaking your overhead settings, these are critical when dealing with small packets.

 

[Authority]

It's not as visible on stock firmware because they do not save the database to disk, it gets wiped every reboot.

Monthly cumulative internet usage gets wipe on every reboot? What am I missing here?

 

[RMerlin]

Monthly cumulative internet usage gets wipe on every reboot? What am I missing here?

Monthly tracking doesn't exist at all on stock firmware (as it would make no sense since the data isn't persistent, being stored only in RAM).

 

[Authority]

Monthly tracking doesn't exist at all on stock firmware (as it would make no sense since the data isn't persistent, being stored only in RAM).

Thanks for the reply. So it's NOT an Asus issue then?

 

[joe scian]

Cake is behaving brilliantly - in my experience much better than FlexQos - at least in what is being reported by DSLReports and fast.com. Difference between unloaded and loaded latency in Fast.com is only 4ms. Thats with
upload and download limits set at 20 and 54 ( my up/down NO QOS speeds are 19 up and 54 down so im actually setting the upload limit higher than the NO QOS upload speed in order to achieve max upload speed). I am getting triple A+ sometimes A for bufferbloat. The download/upload speeds with Cake on those settings is 52.5 download and roughly 18 upload. A perfect addition to our Routers. And thats with no additional Cake configuration settings - just plain Merlin Vanilla.
PS I am on PPPoE VDSL connection 50/20 FTTN ISP plan.
Merlin you are a genius!!

 

[Authority]

Cake is behaving brilliantly - in my experience at least much better than FlexQos - at least in what is being reported by DSLReports and fast.com. Difference between unloaded and loaded latency in Fast.com is only 4ms. Thats with
upload and download limits set at 20 and 54 ( my up/down NO QOS speeds are 19 up and 54 down so im actually setting the upload limit higher than the NO QOS upload speed in order to achieve max upload speed). I am getting triple A+ sometimes A for bufferbloat. The download/upload speeds with Cake on those settings is 52.5 download and roughly 18 upload. A perfect addition to our Routers. And thats with no tinkering with additional Cake configuration settings - just plain Merlin Vanilla.
PS I am on PPPoE VDSL connection 50/20 FTTN ISP connection.
Merlin you are a genius!!

Very jealous, is there a way to use on RT-AC68?

 

[joe scian]

Very jealous, is there a way to use on RT-AC68?

Unfortunately not - why dont you just go out and buy an AC86U - probably not much more than AUS$200 sometimes less on Ebay and retire your existing 68? Best investment you will ever make.

 

[Authority]

Unfortunately not - why dont you just go out and buy an AC86U - probably not much more than AUS$200 sometimes less on Ebay and retire your existing 68? Best investment you will ever make.

I use Ubiquiti myself, but I have an RT-AC68 at my parents that I manage remotely. If I replaced it with anything it would be an EdgeRouter.

TBH they probably don't even NEED QoS... but I'd still use it if it was available. Thanks.

 

[RMerlin]

Thanks for the reply. So it's NOT an Asus issue then?

That's not what I said... All I said is that they do not display monthly data (only daily), and that this data is kept in RAM and therefore vanishes after a reboot, so people using stock firmware almost never pay attention to these. They use Traffic Analyzer data instead.

The issue appeared a few years ago as other people told you, after they started making changes to the rstats daemon to better (?) interface with the network switch data, add VLAN handling, add support for multiple WAN, add 64-bit counter support for newer platforms, and so on. For a while I simply avoided merging their changes in my code to avoid having the same issues, but as of 386 they have done too many changes and fixes to that code for me to no longer keep it in sync with them, and so I merged all of their numerous recent changes with 386.1, which reintroduced the bug in my code.

I have no idea where exactly the bug lies because I have never been able to reproduce it myself, and after visually reviewing the code multiple times I am unable to see where the issue is. The bug will have to be fixed either by Asus (who has the manpower to test/debug this), or someone else who is actually able to reproduce it.

 

[HeadBanger]

I’m going to give Cake QOS a try but I’m struggling to find instructions on what to set the MPU to under the WAN packet overhead - what does this do? Can it be left at the default 0?

Thanks,

HB