What's new

Beta Asuswrt-Merlin 386.2 Beta is now available

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

Status
Not open for further replies.
FWIW, the Cake experiment worked well for me on beta2. I can get 338 down instead of low 200s.
Main problem is whether this will also improve the general experience under other scenarios (like multiple clients running at once, mixture of streamed + latency sensitive applications, etc...). And also whether you need to take into account VLAN tagging when handling traffic from the br0 interface, etc...

Essentially, a lot of variables requiring a lot of testing and in-depth knowledge into tc and trafic scheduling which I lack. And by testing I don't mean running DSLReports' speedtest and calling it good, but actual manual testing, by running multiple types of network loads.

So, I'll be sticking to the most basic setup that is known to work for almost everyone, and leave any finetuning to the community. That is why I spent so much effort in providing flexible "hooks" to allow people to completely customize the way Cake is implemented. Maybe if a few months down the road someone comes with a widely tested architecture that could make sense I might switch implementations, but for the 386.2 release, this will be considered "good enough".
 
I also implemented a potential fix for the rstats traffic spikes. I have no way of testing it, so I have no idea if it resolved that particular issue or not - feedback from users who are able to reproduce the problem would be appreciated.
I don't know if you saw @Vergo 's comment here: https://www.snbforums.com/goto/post?id=674269 , but the comment of interest was:

"BandwidthDetection" is a way of querying the network interface driver for the maximum bandwidth that the interface is capable. If that driver provided information isn't valid then there isn't much vnStat can do. Complain to the manufacturer if you happen to have connections that way. If the driver however opts to provide no information at all then vnStat uses "MaxBandwidth" as fallback value to act as a sanity check for the seen traffic. If the traffic is higher than the given value then it gets ignored. This is used to eliminate "ghost traffic" and other kinds of traffic spikes that may occur due to device driver implementation issues and other reasons. I'd highly recommend not setting "MaxBandwidth" to zero.
So if your new approach doesn't work, perhaps something similar could? (It was a setting that I initially didn't use for vnStat-on-Merlin, because I wasn't clear on what it did, but it may help control those false spikes).
 
So, I'll be sticking to the most basic setup that is known to work for almost everyone, and leave any finetuning to the community. That is why I spent so much effort in providing flexible "hooks" to allow people to completely customize the way Cake is implemented. Maybe if a few months down the road someone comes with a widely tested architecture that could make sense I might switch implementations, but for the 386.2 release, this will be considered "good enough".
I see Cake like a thermostat. You might adjust it frequently as the seasons change, but mostly set it and forget it once you find your comfortable settings.
 
The state of cake on my AX88U with this Beta 2 is awesome @RMerlin . I keep doing things like; run spotify on my phone, download an ubuntu version on my linux computer, played counter strike on my gaming computer, and access one of my cameras all at the same time no problems that I can see. My ISP sells me 60/30 FTTP, I entered 59/29 and an ethernet overhead, keeps impressing me!

EDIT: Forgot to add my TV stream which is always running, I have not tested work from home stuff yet.
 
Last edited:
FWIW, the Cake experiment worked well for me on beta2. I can get 338 down instead of low 200s.

Can confirm, it now limits download with beta2. But I don't see any difference using this.

Also, in my opinion besteffort should be the default for upload instead of diffserv3. But I understand why you would choose diffserv3, a lot of people work from home now and use zoom/teams.
 
Last edited:
I was experiencing an issue where my Honeywell thermostat would not connect to the main 2.4Ghz network after updating to 386.2x. It would connect to the guest network but lose connectivity after a few minutes despite the router listing it as a connected device. I also have a work laptop that was able connect to the guest network but could not establish an internet connection.

Both issues were resolved with Beta 2 and I suspect it was primarily due to the following change.

3b449faf84 rc: revert static DNS routes added with GPL 42095

I don't use many of the other features affected by the changelog, unless they relate to DoT.

Many thanks!

EDIT: I'm using the RT-AC88U
 
So, just rolled out the Beta 2 and I can't seem to connect to 2.4GHz wifi at all. All my Nest/TP-Link Kasa/IOT devices are off-line. 5GHz seems fine. Will roll one back and see.

EDIT: So, rolled back my APs one by one to Beta 1 but still did not alleviate the 2.4GHz issues. Rolled my main router back to Beta 1 and that did the trick. Very strange. I'll try again tonight when there's less family interruption. Thought I could squeeze this one in without them noticing. Doh. Nothing fancy here, only making use of DNSFilter at the router level. APs pretty much vanilla setups. Router running DHCP, Pi-hole's handling DNS.
 
Last edited:
Just about to update on my newly purchased AX88U that I just set up yesterday with beta 1, I expect it will go flawlessly
 
Thanks for the update. Updated both Ax88U.
everything seems to be all connected
Will try cake soon, as I’m running Suricata- not compatible
 
For testing cake performance, here is a test that has been discussed on the Bufferbloat project's email list:


I got an A without any additional tweaking of the 386.2 beta 2 default configuration.

RT-AX58U 150 Mb/s down 20 Mb/s up, cable/docsis. Because Comcast overprovisions bandwidth by ~ 15%, I went w/ 100% of my subscribed values.
 
For testing cake performance, here is a test that has been discussed on the Bufferbloat project's email list:


I got an A without any additional tweaking of the 386.2 beta 2 default configuration.

RT-AX58U 150 Mb/s down 20 Mb/s up, cable/docsis. Because Comcast overprovisions bandwidth by ~ 15%, I went w/ 100% of my subscribed values.

Thank you for this link I think this test will replace the DSL reports for me when anyone asks for a site to test bufferbloat.
 
Installed beta 2 on my ax88. And still internet connection (was not working on beta1)
Going to test it now, will get back later
 
Can confirm, it now limits download with beta2. But I don't see any difference using this.

Also, in my opinion besteffort should be the default for upload instead of diffserv3. But I understand why you would choose diffserv3, a lot of people work from home now and use zoom/teams.

I agree that besteffort should be the default. Zoom, Teams, and every thing else work perfectly with besteffort. The only time diffserv3 would make a difference is if there is an over subscription of bandwidth requiring low priority packets to be dropped. In this situation life is not good no matter what one dose. The only solution is provisioning more bandwidth. The only time cake has work to do is when there are applications that flood the interface such as file transfers. They get slowed by cake only getting slots when all other applications have had all there packets transmitted. Because of this besteffort makes things as simple as cake. All diffserv3 will do on a link that is not oversubscribed is use more CPU.

Morris
 
Merlin: I've seen you state that you can't test Cake. I believe you can by simply setting the upload and download limit to something the processor can handle, for example 200Mb. Of cause you will not get all you are paying for yet you can test and then turn it off. This assumes that others using your network will not start complaining.

Morris
 
For testing cake performance, here is a test that has been discussed on the Bufferbloat project's email list:


I got an A without any additional tweaking of the 386.2 beta 2 default configuration.

RT-AX58U 150 Mb/s down 20 Mb/s up, cable/docsis. Because Comcast overprovisions bandwidth by ~ 15%, I went w/ 100% of my subscribed values.


Amazing test.

Cake on:
cakeOn.jpg


cake off:
cakeOff.jpg
 
Last edited:
This test works much better than the DSLReports test. I love that they provide the actual timings.


Jez, that's amazing, i wish i had results like those. My ISP is trash, I have fiber 10 meters from my house, yet they don't install it.
Stuck on adsl2+. I think i need to try a new modem to see if it lowers my bufferbloat.

What's even worse is that adsl2+ clients are have to pay the same as FTTH clients with 200/100 or 500/100.
 
Jez, that's amazing, i wish i had results like those. My ISP is trash, I have fiber 10 meters from my house, yet they don't install it.
Stuck on adsl2+. I think i need to try a new modem to see if it lowers my bufferbloat.

What's even worse is that adsl2+ clients are have to pay the same as FTTH clients with 200/100 or 500/100.
I'm on Verizon FIOS with an old 100-Mb symmetric limit. They offer much more. I've only had one outage in over 10 years and that was when there router failed. I use there router and thus have double NAT which is OK with me. This way they provide full support for my setup.
 
Status
Not open for further replies.

Latest threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top