[Fork] Asuswrt-Merlin 374.43 LTS releases (Archive)

[dave14305]

I use QoS successfully here on Merlin LTS fork . I established my configuration based on a discussion I read on the Freshtomato-mips forum.

My router is an RT-N66U.

I adjusted the "Up Bandwidth" to 75% of the minimum I observed (after several tests) and the "Down Bandwidth" to 85% of the minimum also observed. It is also these settings that give me (after many tests) the lowest average Bufferbloat results in ms, in the detailed results of "dslreports": http://www.dslreports.com/speedtest/50662106
I have a constant rating of "A+" since I made these adjustments.

List of standards specified by the user:

TCP/UDP, DST port 1-5070, Transferred 0-1kb, class1 (Highest)
any, MAC address (VOIP interface), class2 (High)
TCP/UDP, DST port 1-5070, Transferred 0-64kb, class2 (High)
TCP/UDP, DST port 1-65535, Transferred 0-512kb, class3 (Average)
TCP/UDP, DST port 1-65535, Transferred 0-1024kb, class4 (Low)
TCP/UDP, DST port 1-65535, Transferred 1024kb+ class5 (Lowest)

Download and transmission limits:

1, 15%, 100%
2, 5%, 100%
3, 5%, 100%
4, 5%, 100%
5, 5%, 95% (Set as Default)

(same for both outbound and inbound)

So the more data a connection transfers, the lower its priority?

It may result in good bufferbloat, but what does it do for streaming?

 

[Deleted member 27741]

I think my latency testing situation is a bit different than using something like dslreports.com. Real time latency data in/from online gaming is much more sensitive a test and I have not (after trying about everything imaginable) been able to get QOS to work for latency elimination at that level.

The latency spikes are there no matter what I do with QOS and come from other devices on the network using the internet even though I have tried countless configurations that put my gaming machine on highest and all other traffic low or lowest.

Tourist- your settings look very promising though for small packet latency and I will give them a try!

All that being said, I would still like to like a way to set a limiter that limits all devices OTHER than one machine. I think I might just put that machine on the highest IP address and set all below with a limit as the easiest work-around.

 

[the_tourist]

Video streaming usually uses quite large buffers. Normally it is not affected by short delays to allow small data packets that have priority to pass through.

On the contrary, a VOIP telephone conversation would be unpleasant to listen to, if it were continuously interrupted by large files that should not have priority.

Surfing the web also benefits from being fluid and fast.

And what about the fighting games on the Internet.... Here again the click of a weapon is very small data information... but which needs to have priority for the game to be effective, the large data files of the game are already on the computer disk.

So giving priority to small packages over large ones makes a lot of sense and gives good results.

 

[Deleted member 27741]

Yeah, I will give that a shot for sure.

Meanwhile, I also discovered using a range like 192.168.48.2-192.168.48.253 works in the limiter. With my "unlimited" computer at 192.168.48.254 I will see how that works as well.

Thanks for the advice and double thanks for actually listing settings!!! Very helpful.

 

[the_tourist]

Important:
Remember that QoS is totally incompatible with the "Cut through forwarding" (CTF) option. If you enable CTF, QoS will not work.

 

[ColinTaylor]

If you enable CTF, QoS will not work.

Not quite. If you enable QoS it will disable CTF. There's a subtle but important difference.

 

[Deleted member 27741]

Lucky for me I don't know squat about the CTF option.

Wait, yes I do. My interwebs is far too slow for NAT acceleration to matter, fortunately or unfortunately depending on how you look at it.

UPDATE! I tried your settings and it did not work. QOS will just not eliminate latency spikes for me. Luckily the limiter option works great so I'll just stick to that. I've banged my head on the QOS wall for far too long already.

 

[the_tourist]

"Not quite. If you enable QoS it will disable CTF. There's a subtle but important difference"

Okay, I didn't know that. It's better this way... at least QoS works.

In fact, I wonder why CTF couldn't work with a QoS with these rules, which only checks the size of packets (without analyzing them more than that) before prioritizing them.

 

[dave14305]

QOS will just not eliminate latency spikes for me.

Decrease your download/upload bandwidths until you get something usable. You must ensure the router is the bottleneck by way of these limits.

 

[LupoBiancoU]

Hey guys! I'm pretty new at this stuff and I need some help.

I need to configure my Network to prioritize League of legends (or my PC) as for I don't really play anything else. Have no idea how to set-up, my network has the following:
1 Roku TV (Streaming and downloads when off too)
3 Smartphones (Facebook, Youtube)
2 PC (Gaming)
1 Laptop (Streaming, Netflix, etc)
100/10 Mbps

I know the"I don't need QoS" argument, thing is, with the IMCP and such prioritization game feels so... much... smoother... (My ISP is a huge load of sht and anything (port wise) is good enough to make my gaming experience ruined, feels like no matter what, when anything is using internet thats not my game, game goes to unplayable). I even have to reset my IP every few days because 3 out of 4 IPs given by my ISP give me more ping than the usual. My game withouth anything feels like my ping is jumping all over the place, one click has 50 ms, one click has 100, one click has 40, one ping has 80 but the "ms" at the corner don't move from 55 ms (Weird cause I'm 100% certain it is unplayable).

Right now I have BW Limiter at 70/7 for anything but my PC that isn't capped but, QoS feels better even without the download limit function (I have no idea how to install this driver in my Router and my Router has the 380 FW that has no Download control in traditional QoS).

I've got an RT-N66R (Can't get anything better with dynamic QoS or such, third world country problems).

Any suggestions?

 

[ColinTaylor]

(I have no idea how to install this driver in my Router and my Router has the 380 FW that has no Download control in traditional QoS).

This thread is about John's fork firmware. If you're asking how to install it then just follow the instructions in post #1.

 

[LupoBiancoU]

This thread is about John's fork firmware. If you're asking how to install it then just follow the instructions in post #1.

I asked for a good configuration that could fit my needs because I see everyone sharing theirs, I know how to install it.

 

[ColinTaylor]

, I know how to install it.

That's not what it sounded like:

I have no idea how to install this driver in my Router and my Router has the 380 FW ...

Traditional QoS in Merlin's 380 firmware works differently (i.e. is broken) than in John's fork. If you want to use the examples in this thread you'll need to install this fork.

 

[LupoBiancoU]

An apology, I was trying to expose my problem and the potential solution with this Firmware. I'm in the process of Flashing right now.

 

[Deleted member 27741]

Decrease your download/upload bandwidths until you get something usable. You must ensure the router is the bottleneck by way of these limits.

Thanks for the advice, but I have tried that (I have been tinkering with this issue occasionally for... months). It does not work. Not sure why but QOS does not eliminate latency spiking with anywhere near the results of bandwidth limiting on my N66U. I have not been able to find out why and thanks for your idea it is a good one.

 

[dave14305]

Thanks for the advice, but I have tried that (I have been tinkering with this issue occasionally for... months). It does not work. Not sure why but QOS does not eliminate latency spiking with anywhere near the results of bandwidth limiting on my N66U. I have not been able to find out why and thanks for your idea it is a good one.

One of the resources I read to understand Traditional QoS better was this aged write-up on linksysinfo.org:
Using QOS - Tutorial and discussion

Some things have advanced since then, but the concepts were valuable to me.

 

[Deleted member 27741]

Great link thanks! I was particularly intrigued to see that page suggested 50% of max link speed for preservation of latency! I might have to try that, but at that point I don't see it as much different than using a bandwidth limiter. More granular control for sure, though.

Using the bandwidth limiter to limit "the rest" of the devices on my network to 60% up and 60% down leaves 40% for my latency-sensitive computer tasks and has worked very well so far.


UPDATE! I have tried out QOS with 50% up and 50% down speeds. I set up things so only three devices were on the network and set computer2/3 to stream youtube and tested latency on the main computer. No dice. Note that again, setting a bandwidth limiter at 60% for everything other than my latency-sensitive computer resulted in no latency spikes. Perhaps I am filling out the rules in a way that does not work correctly?! My rules were:

Default priority: lowest

Service name: blank
Source IP: MACofmaincomputer1
Destination port: empty
Protocol: any
Transferred: blank
Priority:highest

Service name: blank
Source IP: MACofcomputer2
Destination port: empty
Protocol: any
Transferred: blank
Priority:lowest

Service name: blank
Source IP: MACofcomputer3
Destination port: empty
Protocol: any
Transferred: blank
Priority:lowest

Upload/download bandwidth settings as you described in your earlier post:
1, 15%, 100%
2, 5%, 100%
3, 5%, 100%
4, 5%, 100%
5, 5%, 95% (Set as Default)

 

[the_tourist]

I would say that your submitted configuration misses all the potential benefits of QoS.

It only reproduces the type of bandwidth distribution you get more simply by bandwidth limiting.

QoS is used to order and prioritize traffic types, not computers.

With rules submitted, even your computer with the highest priority does not benefit from QoS; all its traffic is routed without classification and therefore without being able to order it in such a way as to favour the type of traffic that would benefit from priority. Your computer is the first in this situation not to benefit from the advantages and potential of QoS.

 

[ColinTaylor]

With rules submitted, even your computer with the highest priority does not benefit from QoS; all its traffic is routed without classification and therefore without being able to order it in such a way as to favour the type of traffic that would benefit from priority. Your computer is the first in this situation not to benefit from the advantages and potential of QoS.

This is incorrect. He has created two priority queues, highest and lowest. It's perfectly valid to classify traffic based on its source address rather than something like destination port. This is in contrast to the bandwidth limiter in which all traffic goes through the same queue without prioritisation.

As to why @000111 has his latency problems, I suspect this more to do with his ISP's network than anything else. There's only so much you can do to get around problems that are outside of your LAN.

 

[Deleted member 27741]

Indeed you are right tourist- this is a bit of a "sledgehammer" utilization of QOS that lacks the elegance of your QOS rules (mostly for the sake of a simple/reliable test). However, it should work AT LEAST as well as the bandwidth limiter under these conditions and it does not. To be honest, I don't think it is working at all. For me at least. No idea why.

My other "sledgehammer" solution of setting the ip address of the computer I want to have no latency as the last in the list (254) and limiting everything else with the limiter (2-253) to 60% up/down works well for now. Not sure it is worth chasing down the issue it is more of an academic pursuit since I have a reliable alternative.

I suppose in a few months/years I will be in the market for a new router with better hardware that can handle something like fq_codel.

I must note again (or perhaps for the first time?) that I am using in game metrics (rocket league network graphs and tbh you can see the rubber banding/teleporting/lag easily and it correlates perfectly to the graphs) to measure latency/quality- I am unsure I would ever even see the latency using something like dslreports or other since it comes in spurts (probably caused by video downloads/streaming mostly). Does a jump from 50 to 200 ping matter in daily use for anything other than fps or fast reaction games? Probably not. Does it matter if you biff a save in overtime because you lagged out? Yes, indeed.

Honestly I'm glad I found SOMETHING that works. When others are on their computers in the house I can game latency free. Yay!