I will give this a try later on. Thanks again 
like i said it was to cover
cache1
cache2
it would of been to many to lists in the file.
For anyone that read this ams1 is for the Netherlands.
AMS tends to be shorthand for Amsterdam, which indeed is in the netherlands, but domain names are not guaranteed to be veridical descriptions of anything, let alone the geographical location of something. That said, it is very likely that an ISP/data center operator putting "ams*" into a domain name also places the respective machine there...
2 Likes
@nbd i have upgraded qosfy to 2022-02-10-8d6c19a8 today however i have noticed that CS0 isn't working but CS1 works fine and it seems like the lowest i can go.
Anything i can do to debug it?
qosify
config defaults
list defaults /etc/qosify/*.conf
option dscp_default_tcp bulk
option dscp_default_udp bulk
config class maximum
option ingress CS7
option egress CS7
config class bulk
option ingress CS0
option egress CS0
config interface wan
option name wan
option disabled 0
option bandwidth_up 10mbit
option bandwidth_down 10mbit
option overhead_type pppoe-ptm
# defaults:
option ingress 1
option egress 1
option mode diffserv4
option nat 1
option host_isolate 1
option autorate_ingress 0
option ingress_options ""
option egress_options "wash"
option options "ether-vlan"
config device wandev
option disabled 1
option name wan
option bandwidth 100mbit
00-defaults.conf
tcp:80 maximum
tcp:443 maximum
udp:80 maximum
udp:443 Maximum
From my understanding, anything eles that isn't listed above would get bulk.
If i change this to CS1, The rest of ports would be around 1Mb minimum from a !0Mb download So my understanding CS0 would be atlast roughly be 50-64kbps max if something is downloading from the CS7
What is your config?
Sorry Dave, Forgot to include the info.
How are you concluding that CS0 doesn’t work? What commands?
I am using a linux iso via Torrent to check this for CS0 and then download a test file on port 80.
I think CS0 is broken because it seems to take over and only CS1 seem to throttle the torrent to about 1Mb from a 10Mbit download connection, Again putting CS0 will win the race making port 80 traffic last.
From memory, the November 2021 build of qosify worked with CS0 leaving about 50kbps left of traffic.
Target is a ipq40xx on a fritzbox 7530
Edit:
@dave14305 @nbd sorry, reading Qosify: new package for DSCP marking + cake - #172 by elan explains CS0 is Best Efforts.
Few more DNS for your setup if interested?
# Google Drive Uploads
dns:*.content-storage-upload.googleapis.com bulk
# PlayStation Downloading DNS
dns:*.ps4.update.playstation.net bulk
dns:*.ps5.update.playstation.net bulk
# Gaming DNS
# Call of Duty Lobby/Joining Partys
dns:*.prod.demonware.net gaming
# Tom Clancys Rainbow Six Siege
dns:*.online-services.aws.r6.ubi.com gaming
dns:*.r6.ubi.com gaming
# Xbox Game Studios, Microsoft Studios only (Sea of Thieves, State of Decay, Halo, Forza etc)
dns:*.cloudapp.azure.com gaming
# Discord VoIP Voice max is 128kbps, none server boots are max 96kbps
tcp:50001-50010 telephony
udp:50001-50010 telephony
# TeamSpeak 3 VoiP
udp:9987 telephony
# Mumble VoiP
udp:64738 telephony
1 Like
hi francisuk thanks a lot of i will try tomorow 
hi maybe like this i think no
dns:*demonware.net gaming
1 Like
Have a look at the output of tc -s qdisc, the reported rates are maximum rates for all tins but Bulk, for Bulk this is the guaranteed minimum (for Voice and Video the maximum is also the guaranteed Minimum since these two values add to less than 100%).
CS0 ends up in Best Effort and that will get up to almost everything that is not used by Voice or Video (or the Bulk guaranteed minimum), but note the combined max rates for voice and video and the minimum for bulk add up to less than 100%, so Best Effort will always get some packages sent, at worst:
MinBestEffort% = 100% - MaxVoice% - MaxVideo% - MinBulk%
I think your reported results, as far as I understand, are fully within the expected, no?
1 Like
@moeller0 This post helped me, Qosify: new package for DSCP marking + cake - #172 by elan as i thought CS0 was lower then CS1.
Setting CS1 as bulk on a 10Mb Download, leaves the following
CS7 maxed out will leave 3Mb
CS4 maxed out will leave 1Mb
seems to work good.
dns:*demonware.net gaming
Normally there somethinghere.prod.demonware.net
1 Like
@francisuk1989 cs1 is bulk
yes this post of elan is very important and we are with elan test configure for gaming
1 Like
"Change toolchain option to use a prebuilt LLVM toolchain, and enable qosify."
I'm lost here. How we can do that?
Use make menuconfig, or edit the .config by hand.
I only added the two above lines to my pruned .config recipe ( which I expand to the full .config with "make defconfig")
1 Like
Hello guys. I am a noob here. I come after trying to read if this was discussed but I am not sure. After configuring Qosify with the help of all your previous posts, I think it works amazingly well to prioritize certain tasks such as gaming. However, I come here to ask a very crazy question: could I get the best of both worlds with Qosify and Fq_codel (that is to have DSCP marking and traffic shaping)?
I ask this because I am using a Belkin RT3200 and I have a gigabit connection and with Fq_codel I can get all the juice from it, but with Cake I can only get half of it at most. I tried to set my configuration with the bandwidth options option bandwidth_up 0 and option bandwidth_down 0, and of course commenting out any type of overhead option as well. Then, I enabled Fq_codel under Sqm Qos and so far, it seems that it could be working or at least just using fq_codel. I hope that someone smarter than I can validate or destroy my dreams, lol.
I apologize in advance if a similar question has been asked before, or if my assumptions are just too wrong to begin with, lol. Thanks!
1 Like
One last question what is the buildroot location so that I can place the file in order to extract in that directory
No need.
Qosify already launches the cake qdisc, so you do already have traffic shaping. Fq_codel is an alternative qdisc to cake...
"SQM" is just a qdisc launcher/configurer script, and the shaping is done by the underlying qdisc, cake or fq_codel as the main alternatives.
Qosify adds the selective prioritization marking layer, so that the cake qdisc can better prioritize traffic according to your wishes.
But there is no need to run SQM in addition to qosify on the same interface. (With more complex routing, there might be use cases of having them in different interfaces.)
1 Like
Buildroot location ???
The practical root of your cloned OpenWrt source repo and toolchain, the dir from which you issue all build commands etc. The .config file also is there.
1 Like
Cake is more CPU intensive, as it classifies more. With gigabit traffic your router CPU can get choked.
If you set up and down limits to 0, meaning unlimited, what is the point of fq_codel?
The main point of QoS tools is the ability to smoothly and selectively limit traffic so that the latency remains small and you avoid bufferbloat.
With gigabit connection, your actual need for traffic limitation is usually small. Not sure if you really benefit from qosify if also your upload bandwidth is wide.
2 Likes
hello. Sorry, I didn’t explain properly. I set the configuration services for bandwidth in Qosify with zeros and then I enabled proper settings in Sqm using fq_codel. That way (hopefully) I get the DSCP markings from Qosify, but the traffic shaping from fq_codel. That way, I could shape the gigabit download. Unfortunately, I do need some traffic shaping because my connection is docsis and kind of bad with only 50 mbps up.
2 Likes
With something so asymmetric as 1000:50, you might be best off by not shaping the incoming download, but shaping the narrow upload at 47 or so. Then you avoid wasting CPU power for the massive download, but benefit from managed latency in upload channel. (I doubt that you really have a constant incoming traffic at gigabit level outside the speed tests...)
1 Like