I am curious if fq_codel is available in these latest beta releases?
I would’t think so. Support for it before Linux 3.5 is nonexistent and even after 3.5 it is still very sparse. It seems that 3.12 is the lowest sensible requirement. Can we expect it to land in v7? Normis?
oh yeah. fixing bloatbuffering and recent kernels used in RouterOS, thus - would be LOT of help.
but seems they presently focused on polishing Tilera code bit more than moving upstream.
which is sane.
I too would be quite interested in this.
i think its matter of time.
cuz newer kernels bring LOT of different improvements(aside Co_Del and other measures against blotbuffer, incrimental nearly in ANY next/new release), quite benefiticary for any kind of devices.
fq_codel support would be amazing and could really improve our experience.
(Unfortunately, this seems almost entirely based on the kernel version.)
To clear up some stuff - fq_codel became stable in linux 3.6, and was backported by qualcomm-atheros folk as far back as linux 2.6.32 as part of the compat-wireless patchset. If you are using that patchset on an older kernel, you already have fq_codel.
Most of the core de-bufferbloating work entered the kernel before linux 3.10, and linux 3.10.12 and later (there were some bugs in htb backported from 3.11) have everything that is “core” to a router’s needs for vastly better queue management. Now that 3.10 is stable (and the list of products using fq_codel other than mikrotik is getting rather long), some new work is starting up in the 3.18 series that will be interesting.
Useful stuff to have in general are bql support on the ethernet and dsl device drivers, also. That is generally a tiny amount of work per bql-enabled device driver for a large latency improvement at low speeds, and a noticible one at all line rate speeds.
A frustrating aspect of most of the new router designs is that much offloads are buried in the network hardware to make it “go fast”, but they are difficult to turn off when it is needed to slow down enough to manage a slow uplink (eg, ISP gateway) more intelligently. The cavium used in the edgerouter is a classic example of this problem. Using software rate limiting to manage an ISP link peaks out at about 60Mbit. I’d hope that the tilera stuff in particular, would have either enough oomph to do software rate limiting better, or be able to embed algos such as htb + fq_codel in their offload engines in general.
An example of the benefit of fq_codel (and the above problem) is here:
http://burntchrome.blogspot.com/2014/08/new-comcast-speeds-new-cerowrt-sqm.html
At lower speeds the bufferbloat problem is much worse, and most hardware can handle these compensating mechanisms without raising a sweat, if it has the core stuff available.
I am (personally) in search of a routing product that can handle up to 300mbit software rate limiting + fq_codel…
Looks like there was a discussion this summer about CoDel here: http://forum.mikrotik.com/t/codel-support/58063/1
We’ve been testing this on some saturated servers and other *nix routers with great results. Hoping to see it in v7.
fq_codel is really good thing for router or heavily loaded server.
Please add support of it into RouterOS.
It works well with HTB tree (as good as RED or SFQ).
It was ported to lot of kernels so it can be in RouterOS too.
Hope it can be done easily enough and you will add it soon 
thanks for the suggestion, we are looking into it for v7. currently you can use SFQ, which is also very good
Hi,
I know, this is an old thread, but I have a question to the last thread posted by normis.
you said, instead of using fq_codel, you can use SFQ and this also works quite well.
I was looking for some HowTo this, but unfortunately, I did not find anything about SFQ and reducing bufferbloat.
Can you give me some advices or maybe a short HowTo? 
My bandwidth is 30/6 Down/Up Mbit.
With a full upload (about 5,8Mbit), my ping times rise from normal 6ms to my ISP nameserver to 250-900ms.
Regards,
Florian
I use queue trees for this(without SFQ as it slows down things for me).
I got 3-8ms jump at full download or upload (which is more than 500mbit).
If you wish i can post my queue rules.
Hi PtDragon,
would be great, when you can post your rules!
Hopefully, I can rewrite these rules to use them in my config, but I will do my best!
THX,
Florian
I attached full export of /ip/firewall/mangle and /queue
I can clean up that mess a bit if you need just for one interface.
Hi,
thank you very much! I looked into your config and I have to say, that you have a lot of WAN interfaces! 
I tried to separate one interface from both config files, but mangle config is quite hard to understand with so many interfaces and rules.
The next thing, I don’t understand is that you have in queue config, different parent interface for ethX-IN and ethx-OUT.
Another thing is also not completly clear to me, which rules from mange config, I really need for the queue part.
It would be great, if you can shrink down your config to only one WAN interface.
Do you have 100Mbit Up/Downlink? If yes, why do you use only 95Mbit for rate limit?
Regards
Florian
I looked into your config and I have to say, that you have a lot of WAN interfaces! >
Yeah, that’s true. I have 6 links from 3 ISPs right now.
The next thing, I don’t understand is that you have in queue config, different parent interface for ethX-IN and ethx-OUT.
It’s because I limit both incoming and outgoing traffic for each external interface.
I need to limit it to avoid big delays on ISP network.
Do you have 100Mbit Up/Downlink? If yes, why do you use only 95Mbit for rate limit?
I have 100Mbit on both interfaces but limit to 95 to avoid delays inside ISP limiting queues(which varies from 20 to 300ms).
Not reaching limit = no additional delay from limiting by ISP.
It would be great, if you can shrink down your config to only one WAN interface.
Sure, I will do it today or tonight.
Sorry guys, was a bit busy(when working on several jobs at same time it happens 
)
queue:
/queue tree
add limit-at=1 max-limit=95M name=ETH1-IN parent=bonding1 queue=RED16
add limit-at=1 max-limit=95M name=ETH1-OUT parent=ether1 queue=RED16
add name=ETH1-IN-1 parent=ETH1-IN priority=1 queue=PFIFO_8
add name=ETH1-IN-4 parent=ETH1-IN priority=4 queue=PFIFO_8
add name=ETH1-IN-8 parent=ETH1-IN queue=PFIFO_8
add name=ETH1-IN_BT packet-mark=wan0_bt_in parent=ETH1-IN-8 queue=RED16
add name=ETH1-OUT-1 parent=ETH1-OUT priority=1 queue=PFIFO_8
add name=ETH1-OUT-4 parent=ETH1-OUT priority=4 queue=PFIFO_8
add name=ETH1-OUT-8 parent=ETH1-OUT queue=PFIFO_8
add name=ETH1-OUT_DNS packet-mark=wan0_dns parent=ETH1-OUT-1 queue=PFIFO_32
add name=ETH1-OUT_ICMP packet-mark=wan0_icmp parent=ETH1-OUT-1 priority=4 queue=PFIFO_8
add name=ETH1-OUT_HTTP packet-mark=wan0_http parent=ETH1-OUT-4 priority=4 queue=PFIFO_32
add name=ETH1-OUT_SSH packet-mark=wan0_ssh parent=ETH1-OUT-4 priority=1 queue=PFIFO_8
add name=ETH1-OUT_SSL packet-mark=wan0_ssl parent=ETH1-OUT-4 priority=3 queue=PFIFO_32
add name=ETH1-OUT_BT packet-mark=wan0_bt parent=ETH1-OUT-8 queue=RED16
add name=ETH1-OUT_SSL_B packet-mark=wan0_ssl_b parent=ETH1-OUT-4 priority=6 queue=RED16
add name=ETH1-IN-4-1 parent=ETH1-IN-4 priority=1 queue=PFIFO_8
add name=ETH1-IN-4-8 parent=ETH1-IN-4 queue=PFIFO_8
add name=ETH1-IN-2 parent=ETH1-IN priority=2 queue=PFIFO_8
add name=ETH1-OUT_OTHER packet-mark=wan0p_out parent=ETH1-OUT-8 queue=RED16
add name=ETH1-OUT_ACK packet-mark=ack_wan0 parent=ETH1-OUT-1 priority=1 queue=RED16
add name=ETH1-IN_OTHER packet-mark=wan0p_in parent=ETH1-IN-8 priority=4 queue=RED16
add name=ETH1-IN_ACK packet-mark=wan0_ack_in parent=ETH1-IN-1 priority=1 queue=RED16
add name=ETH1-IN_ICMP packet-mark=wan0_icmp_in parent=ETH1-IN-2 priority=4 queue=PFIFO_8
add name=ETH1-IN_DNS packet-mark=wan0_dns_in parent=ETH1-IN-2 queue=PFIFO_32
add name=ETH1-IN_HTTP packet-mark=wan0_http_in parent=ETH1-IN-4-1 priority=4 queue=PFIFO_32
add name=ETH1-IN_SSH packet-mark=wan0_ssh_in parent=ETH1-IN-4-1 priority=1 queue=PFIFO_8
add name=ETH1-IN_SSL packet-mark=wan0_ssl_in parent=ETH1-IN-4-1 priority=3 queue=RED16
add name=ETH1-IN_SSL_B packet-mark=wan0_ssl_in_b parent=ETH1-IN-4-8 priority= 3 queue=RED16
95Mbit is 95% of bandwidth limit on interface(100Mbit).
Queue types:
PFIFO_numpackets=simple pfifo queue, number is number of packets
REDnumpackets=red queue, number is number of packets.
Avg packet size=1000.
Min threshold=25%of size.
Max threshold=75% of size.
Burst=50%of size.
Firewall part:
/ip firewall mangle
add action=set-priority chain=prerouting new-priority=from-dscp
add action=set-priority chain=prerouting new-priority=from-ingress
add action=mark-connection chain=prerouting connection-state=new dst-address-list=LAN new-connection-mark=LOCAL src-address-list=LAN
add action=fasttrack-connection chain=prerouting connection-mark=LOCAL
add action=jump chain=input connection-state=new in-interface=!bonding1 jump-target=mark_in_only
add action=mark-connection chain=mark_in_only in-interface=ether1 new-connection-mark=wan0
add action=return chain=mark_in_only
add action=jump chain=prerouting connection-mark=no-mark connection-state=new in-interface=!bonding1 jump-target=mark_conn_input
add action=mark-connection chain=mark_conn_input connection-mark=no-mark in-interface=ether1 new-connection-mark=wan0
add action=return chain=mark_conn_input
add action=jump chain=prerouting comment="Jump to prerouting connection marks (PCC)" connection-mark=no-mark connection-state=new dst-address-list=!LAN in-interface=bonding1 jump-target=mark_prerouting_internal
add action=mark-connection chain=mark_prerouting_internal comment=wan0 new-connection-mark=wan0 nth=6,1
add action=return chain=mark_prerouting_internal
add action=jump chain=prerouting comment="Jump to prerouting connection marks (PCC)" connection-mark=no-mark connection-state=new in-interface=bonding1 jump-target=mark_prerouting_internal
add action=jump chain=prerouting comment="Jump to prerouting connection marks (PCC)" connection-mark=no-mark connection-state=new in-interface=bonding1 jump-target=mark_prerouting_internal
add action=jump chain=prerouting comment="Jump to force_marks" connection-mark=no-mark connection-state=new dst-address-list=!LAN in-interface=bonding1 jump-target=mark_prerouting_internal_f
add action=jump chain=output connection-state=new jump-target=mark_out out-interface=!bonding1
add action=mark-connection chain=mark_out new-connection-mark=wan0 out-interface=ether1
add action=return chain=mark_out
add action=jump chain=prerouting comment="Jump to mark routing prerouting" connection-mark=!no-mark in-interface=bonding1 jump-target=mark_routing_prerouting
add action=mark-routing chain=mark_routing_prerouting connection-mark=wan0 new-routing-mark=wan0
add action=return chain=mark_routing_prerouting
add action=jump chain=prerouting comment="Marking incomming packets" jump-target=mark_prerouting
add action=mark-packet chain=mark_prerouting in-interface=ether1 new-packet-mark=wan0p_in
add action=mark-packet chain=mark_prerouting in-interface=bonding1 new-packet-mark=bonding_in
add action=return chain=mark_prerouting
add action=jump chain=forward comment="Marking WAN output packets" jump-target=mark_wan out-interface=!bonding1
add action=mark-packet chain=mark_wan new-packet-mark=wan0p_out out-interface=ether1
add action=return chain=mark_wan
add action=jump chain=forward in-interface=ether1 jump-target=wan0_proto_in out-interface=!ether1
add action=jump chain=forward in-interface=!ether1 jump-target=wan0_proto out-interface=ether1
add action=mark-packet chain=wan0_proto dst-limit=400,100,addresses-and-dst-port/10ms layer7-protocol=http new-packet-mark=wan0_http
add action=mark-packet chain=wan0_proto dst-limit=800,200,addresses-and-dst-port/10ms layer7-protocol=http new-packet-mark=wan0_http_b packet-mark=wan0p_out
add action=mark-packet chain=wan0_proto new-packet-mark=wan0_icmp protocol=icmp
add action=mark-packet chain=wan0_proto layer7-protocol=bittorrent new-packet-mark=wan0_bt
add action=mark-packet chain=wan0_proto layer7-protocol="5TP" new-packet-mark=wan0_bt
add action=mark-packet chain=wan0_proto new-packet-mark=wan0_bt p2p=bit-torrent
add action=mark-packet chain=wan0_proto new-packet-mark=wan0_bt protocol=tcp src-address=192.168.4.64 src-port=16508
add action=mark-packet chain=wan0_proto new-packet-mark=wan0_bt protocol=udp src-address=192.168.4.64 src-port=16508
add action=mark-packet chain=wan0_proto dst-limit=800,200,addresses-and-dst-port/10ms layer7-protocol=ssl new-packet-mark=wan0_ssl packet-mark=wan0p_out
add action=mark-packet chain=wan0_proto layer7-protocol=ssl new-packet-mark=wan0_ssl_b packet-mark=wan0p_out
add action=mark-packet chain=wan0_proto layer7-protocol=ssh new-packet-mark=wan0_ssh packet-mark=wan0p_out
add action=mark-packet chain=wan0_proto dst-limit=800,200,addresses-and-dst-port/10ms layer7-protocol=ssh new-packet-mark=wan0_ssh_b packet-mark=wan0p_out
add action=mark-packet chain=wan0_proto layer7-protocol=dns new-packet-mark=wan0_dns
add action=return chain=wan0_proto
add action=mark-packet chain=wan0_proto_in dst-limit=400,100,addresses-and-dst-port/10ms layer7-protocol=http new-packet-mark=wan0_http_in packet-mark=wan0p_in
add action=mark-packet chain=wan0_proto_in layer7-protocol=http new-packet-mark=wan0_http_in_b packet-mark=wan0p_in
add action=mark-packet chain=wan0_proto_in new-packet-mark=wan0_icmp_in protocol=icmp
add action=mark-packet chain=wan0_proto_in layer7-protocol=bittorrent new-packet-mark=wan0_bt_in
add action=mark-packet chain=wan0_proto_in layer7-protocol="5TP" new-packet-mark=wan0_bt_in
add action=mark-packet chain=wan0_proto_in new-packet-mark=wan0_bt_in p2p=bit-torrent
add action=mark-packet chain=wan0_proto_in dst-address=192.168.4.64 dst-port=16508 new-packet-mark=wan0_bt_in protocol=tcp
add action=mark-packet chain=wan0_proto_in dst-address=192.168.4.64 dst-address-list=local_ips dst-port=16508 new-packet-mark=wan0_bt_in protocol=udp
add action=mark-packet chain=wan0_proto_in dst-limit=200,100,addresses-and-dst-port/100ms layer7-protocol=ssl new-packet-mark=wan0_ssl_in packet-mark=wan0p_in
add action=mark-packet chain=wan0_proto_in layer7-protocol=ssl new-packet-mark=wan0_ssl_in_b packet-mark=wan0p_in
add action=mark-packet chain=wan0_proto_in layer7-protocol=ssh new-packet-mark=wan0_ssh_in
add action=mark-packet chain=wan0_proto_in layer7-protocol=dns new-packet-mark=wan0_dns_in
add action=return chain=wan0_proto_in
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=ack
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=fin,rst,ack
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=fin,syn,ack
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=syn
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=rst
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=rst,ack
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=fin,ack
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=syn,ack
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=0-128 protocol=tcp tcp-flags=fin,ack
add action=jump chain=forward jump-target=mark_ack_wan out-interface=!bonding1 packet-size=129-809 protocol=tcp tcp-flags=ack
add action=mark-packet chain=mark_ack_wan connection-mark=wan0 new-packet-mark=ack_wan0 out-interface=ether1
add action=return chain=mark_ack_wan
add action=jump chain=prerouting in-interface=!bonding1 jump-target=mark_ack_in_wan packet-size=0-128 protocol=tcp tcp-flags=ack
add action=mark-packet chain=mark_ack_in_wan connection-mark=wan0 in-interface=ether1 new-packet-mark=wan0_ack_in
add action=return chain=mark_ack_in_wan
add action=change-dscp chain=postrouting comment="Setting DSCP for the switch" new-dscp=38 out-interface=bonding1
add action=change-dscp chain=postrouting new-dscp=46 out-interface=bonding1 packet-size=0-128 protocol=tcp tcp-flags=ack
add action=change-dscp chain=forward comment="Setting DSCP for the ISP" new-dscp=38 out-interface=!bonding1
add action=change-dscp chain=forward new-dscp=46 packet-mark=ack_wan0
You should use your marks instead of NTH or PCC for your config!
Internal interface is bonding1
External interface is ether1
LAN address list is for internal LAN routing on bonding1(which can’t be balanced to external interfaces so marked before everything else).
Config isn’t perfect and is designed for fast routers such as CCR1036 and may slow things down on routers with not enough CPU power(I didn’t test on other routers).
things like RRED https://en.wikipedia.org/wiki/Robust_random_early_detection
ECB support and DCTCP support would help a lot potentially along with native MPTCP(multi-path TCP) support(already released 0.90 version which is cool)
Just putting my vote in for this. CoDel would be a very welcome addition to the QoS options in RouterOS.