How to Mangle, Priority with HTB (Queue Tree) in ROS 3.16

[samsoft08]

mknnoc , when the test-user hit the 128k , there will be no room for any traffic even it has the highest priority , so priority will work when you dont really need it , and it doesnt work when you need it .. actually nothing will help when test-user hit the 128k , this is an important principle i’d learn here !!! you can check another topic about priority to confirm this fact..

if ICMP is your problem so : isolate it , exclude it by mangle , give it a simple unlimited queue , put this q always on top of other simple q’s , it wont take much bandwidth at all ..

[mknnoc]

Thanks Samsoft for your suggestion! i will use it for my last resort.

anyway, here corrected and tested configuration:

ip firewall mangle print

136 ;;; icmp-packet-upload
chain=prerouting action=mark-packet new-packet-mark=icmp-packet-upload passthrough=no protocol=icmp in-interface=vlan5-noc

137 ;;; icmp-packet-download
chain=prerouting action=mark-packet new-packet-mark=icmp-packet-download passthrough=no protocol=icmp

138 ;;; test-conn
chain=prerouting action=mark-connection new-connection-mark=test-conn passthrough=yes src-address=172.16.31.89

139 ;;; test-packet-upload
chain=prerouting action=mark-packet new-packet-mark=test-packet-upload passthrough=no in-interface=vlan5-noc connection-mark=test-conn

140 ;;; test-packet-download
chain=prerouting action=mark-packet new-packet-mark=test-packet-download passthrough=no connection-mark=test-conn

queue tree print

120 name=“test-parent-upload” parent=global-in packet-mark=“” limit-at=128000 queue=default priority=8 max-limit=100000 burst-limit=0 burst-threshold=0 burst-time=0s

121 name=“test-parent-download” parent=global-in packet-mark=“” limit-at=128000 queue=default priority=8 max-limit=1000000 burst-limit=0 burst-threshold=0 burst-time=0s

122 name=“icmp-priority-download” parent=test-parent-download packet-mark=icmp-packet-download limit-at=128000 queue=default priority=2 max-limit=128000 burst-limit=0 burst-threshold=0 burst-time=0s

118 name=“icmp-priority-upload” parent=test-parent-upload packet-mark=icmp-packet-upload limit-at=128000 queue=default priority=2 max-limit=128000 burst-limit=0 burst-threshold=0 burst-time=0s

119 name=“test-user-download” parent=test-parent-download packet-mark=test-packet-download limit-at=128000 queue=defaul priority=8 max-limit=128000 burst-limit=0 burst-threshold=0 burst-time=0s

0 name=“test-user-upload” parent=test-parent-upload packet-mark=test-packet-upload limit-at=128000 queue=default priority=8 max-limit=256000 burst-limit=0 burst-threshold=0 burst-time=0s

queue type print

13 name=“pcq-128k-up” kind=pcq pcq-rate=256000 pcq-limit=50 pcq-classifier=src-address pcq-total-limit=5000

14 name=“pcq-128k-down” kind=pcq pcq-rate=128000 pcq-limit=50 pcq-classifier=dst-address pcq-total-limit=5000

My testing procedures are:

1. try to use full bandwidth
   result: all are good. ICMP can get good result.
2. try to disable(#122,#188) priority and shapping of ICMP packet, and use full bandwidth
   result: user can get the top speed (D:128k,U:256k) BUT ICMP still good. i think mangle excluded ICMP from user connection already as Samsoft said

if ICMP is your problem so : isolate it , exclude it by mangle , give it a simple unlimited queue , put this q always on top of other simple q’s , it wont take much bandwidth at all ..

3. try to change from ICMP to HTTP protocol. just to make sure that user cannot use over the bandwidth as Janisk said

This way we get virtually the same behaviour as “per user prioritization”

Here is new configuration:

ip firewall mangle print

136 ;;; http-packet-upload
chain=prerouting action=mark-packet new-packet-mark=icmp-packet-upload passthrough=no protocol=tcp in-interface=vlan5-noc dst-port=80

137 ;;; http-packet-download
chain=prerouting action=mark-packet new-packet-mark=icmp-packet-download passthrough=no protocol=tcp dst-port=80

queue tree print

118 name=“http-priority-upload” parent=test-parent-upload packet-mark=icmp-packet-upload limit-at=128000 queue=default priority=2 max-limit=1000000 burst-limit=0 burst-threshold=0 burst-time=0s

122 name=“http-priority-download” parent=test-parent-download packet-mark=icmp-packet-download limit-at=128000 queue=default priority=2 max-limit=1000000 burst-limit=0 burst-threshold=0 burst-time=0s
result: user can use over bandwidth. that mean, PCQ can not help.

[mknnoc]

anyone please??

[mknnoc]

i am waiting for the answer or suggestion. anyone???

[dot-bot]

Here’s the answer my friend. Set queue=pcq-*** to all childs. Adjust it according to upload and according to download - to separate queue types. Then PCQ will work like a charm. Tested and will probably use it for the future like this.

[mknnoc]

Sorry that make you confuse about my configuration. actually, i have been used queue=pcq-*** for test-user-upload and test-user-download. i pasted wrong config, which is queue=default, to the post.

however, i have tested with queue=pcq-***. but customer still can uses over bandwidth if i do HTTP priority.

[dot-bot]

The reason is because due to HTB structure and queue lengths (pcq-limit, pcq-total-limit) the traffic which gets more bandwidth is able to push more packets. I had the same problem with my dedicated p2p machine. Remarked it’s traffic and problem was solved.

In your case you must be more careful to adjust pcq-limit and pcq-total-limit. Or you could separate QoS from PCQ. Leave this setup for QoS and raise all limit-at values. Add another marking and another Queue Tree for PCQ.

@normis

QoS includes several facilities, in the following order:

1. Mangle chain prerouting
2. HTB global-in
3. Mangle chain forward
4. Mangle chain postrouting
5. HTB global-out
6. HTB out interface

So, in one router, you can shape twice if you use:

a) #1 and #2 for first marking and shaping, and #3+#5 for second
b) #1 and #2 for first marking and shaping, and #3+#6 for second
c) #1 and #2 for first marking and shaping, and #4+#5 for second
d) #1 and #2 for first marking and shaping, and #4+#6 for second

@janisk

In my presentation I told that creating priorities separately for each client is suicide - there is no hardware that can handle small queue tree for every user (if you have 1000 of them). So in my presentation I discuss next best thing, which is close as possible to desired behavior.

The main Idea of the setup is to have two separate QoS steps:

1. In the first step we prioritize traffic, we are making sure that traffic with higher priority has more chance to get to the customers than traffic with lower priority.

Example:
We have total of 100Mbps available, but clients at this particular moment would like to receive 10Mbps of Priority=1 traffic 20Mbps of Priority=4 and 150Mbps of Priority=8.
Of course after our prioritization and limitation 80Mbps of priority=8 will be dropped. And only 100Mbps will get to the next step.

2. Next step is per-user limitation, we already have only higher priority traffic, but now we must make sure that some user will not overuse it, so we have PCQ with limits.

This way we get virtually the same behavior as “per user prioritization”.

@Macgaiver

So the plan for you might be to mark by traffic type in prerouting and limit by traffic type in global-in.
Then remark traffic by IP addresses in forward and limit them on the outgoing interface.

1. you need to mark all traffic at the same place (prerouting) (that would be shaped by one Queue)
2. you must mark upload and download for every type of traffic separately (if you use global-total or Queue Simple)
3. you must have a parent queue, that has max-limit and parent=global-in - all other queues parent=
4. you need 2 sets of those queues - one for upload one for download

Priority doesn’t work without limitation.

@janism.pdf

Each simple queue creates 3 separate queues:
One in global-in (“direct” part)
One in Global-out (“reverse” part)
One in Global-total (“total” part)
Simple queues are ordered - similar to firewall rules
further down = longer packet processing
further down = smaller chance to get traffic
(necessary to reduce number of queues)

Queuing Placement
Limitation for in mangle chain “forward” marked traffic can be placed in the “global-out” or interface queue

If queues will be placed in the interface queues
queues on the public interface will capture only client upload
queues on the local interface will capture only client’s download

If queues will be placed in global-out
download and upload will be limited together (separate marks needed)

[dot-bot]

@Equis

I shape Inbound Traffic
The key is to set your upper limit about 15% lower than your service can push.
so if I had 10Mbps from my supplier I would set my top to 9Mbps.

@gustkiller

i think if you just do the queue tree and the MAIN QUEUE for upload and the MAIN QUEUE for download for an example 95% of your total down/up bandwidth the priorities and qos should work better coz they will have some room to work.

[macgaiver]

@gustkiller

i think if you just do the queue tree and the MAIN QUEUE for upload and the MAIN QUEUE for download for an example 95% of your total down/up bandwidth the priorities and qos should work better coz they will have some room to work.

Take a look at:
http://wiki.mikrotik.com/wiki/HTB
to be more precise at:
http://wiki.mikrotik.com/wiki/Image:Stage1_1.jpg

Queue01 have max-limit=50M and have 2 children Queue02 and Oueue08 with max-limit=49M
Queue08 have max-limit=49M and have 2 children Queue09 and Oueue12 with max-limit=47M
Queue09 have max-limit=47M and have 2 children Queue10 and Oueue11 with max-limit=45M

This way if one child queue use all available traffic, other child queue will have a “room to work”. this I was trying to explain by:

@macgaiver

Imagine you are in metro train and in the next station granny is waiting.
Granny is highest priority and as soon as she will get into train someone for sure will give her a place to sit. (at least where I’m from)
Now imagine that train is so full, that granny will not even get into train, so at this point it doesn’t matter what priority granny had, she is staying on the station.
But if train was only 95% full granny will have no problems to get in and use her highest priority.

and this

@macgaiver

Imagine high-speed road with lots of traffic on it, and there are one car coming from side road and need to get on the main high-speed road. In order to make it much more easier you need an acceleration line - where car can build up speed to main roads speed and then blend into rest of the traffic. If there are no acceleration line, this car will have hard time getting on the main road.

If there are 100% of traffic used - similar to situation without acceleration line
If there are 95% and less of traffic used - similar to situation with acceleration line

[mknnoc]

As i know so far, Priority will work only if bandwidth is full(100%). isn’t it?

[dot-bot]

Packet rearrangement occurs in a well built HTB/Queue. Priorities determine which child leaf gets the available limit-at/max-limit of it’s parent. First prio 1 gets it’s traffic, then 2, then 3 etc to 8 for limit-at. After this is done first prio 1 gets for max-lmit, then prio 2 gets for max-limit etc.

[NetworkPro]

You’re kinda wrong. Packets of one connection can not/must not get rearranged, their order must be preserved. We are talking about bandwidth here and packet discarding, not moving forward and backward… I am confused

[Chupaka]

well, packet rearrangement inside a single connection is normal, it’s what TCP was developed for =) but in normal situation packets from one connection should not have different priorities, so they won’t be rearranged

[newtoCS]

Could anyone answer to my post?
http://forum.mikrotik.com/t/where-should-i-implement-htb/39455/1

Please help…