Burst for each TCP connection

RouterOS General
21

Ok. Thx mate!

22

cool code dude…

but if i had 3 different client group in my office… each group the ip listed in adrress list.
ie:
ip1 192.168.1.1
ip1 192.168.1.5
ip1 192.168.1.7
ip2 192.168.1.4
ip2 192.168.1.12
ip2 192.168.1.2
ip3 192.168.1.10
ip3 192.168.1.11
group 1 limit traffic by 96 kbps download 96 kbps upload
group 2 limit traffic by 128 kbps download 64 upload
group 3 limit traffic by 200 kbps download 128 upload

each group using PCQ Equal Share with mangle and queue tree writen in wiki
how to implemented those script to my script?

so i must write 2 connection mark? and 2 queue tree for different group, is that right?

another question.
i had idea to traffic prioritizing in each group, which email traffic (pop3)to be higher priority than http traffic. how to implemented like these to PCQ Equal Share in wiki???

23

Dear Friends,

I have tested in my network and there is a nice problem :wink: that may be reason you to dont use this code or you may find the way to solve this problem.

In my situation which i have set the 1024kbps for first 500k and then fall down to 512kbps for special users when a user download a big file after 2 or 3 MB fall down to 512kbps BUT if that person use software like IDM (Internet Download Manager) each time he/she click on STOP and then again START he/she will get 1024kbps speed :smiley:
ofcouse no one call me to tell me because i didnt use this code generally just for test only and I found this :open_mouth:

Please tell me your opinion about this. it wil be nice if we could find the way to block this kind of trick which users may (actually) will use.

24

Dear friend Antixx,

Yes there is a problem with download managers, cause if clients stops and starts his/her download after every 1mb then it’s starts as a freash (1st rule) rate ratio. and with download manager there has few options like 5 connections for a file download that’s why the download bandwidth fluctuate very time between 1st and 2nd pcq rule. you can close per IP connection limites with firewall.

Hope you understand that

Thanks

25

I think i understand the problem and the reason. but in my view if i limit the connected to each dst ip to 5 connection this problem will occure again? isnt it?

26

if there is wackos to stop start download every minute or so, then let them do that . . .

27

Please how can this be applied on a mt with load balancing?
Thanks

28

first i would see that user that stops download after each X Mbit you have set in a download.

please read manual about simple queues and queue tree. if you have several groups, then just add tree nodes to the queue tree and set up configuration for them, so, they function properly, you will have to mark connections and then packets for each type of users, you just add dst-address-list for download and src-address-list for upload for each group.

after that you simple follow the example

29

i think it is misunderstanding, with this setup usercan achive more bandwidth by doing stop and start the download manager software, anyway to solve this problem?

it will be great doing this…

Thank you very much in did/

30

Hi to all. I’ve tested this configuration on my MT 2.9.51 at work. And i must say is wonderful.But the problem is if i start to download with flashget, configured to download with only 1 connections, this setup is perfect.Sow when i configured flashget to download with 5-10 connections, the download speed vary from 1st rule to the 2nd rule after 1 min and after 3 min vary from 2nd rule to the 1st, and so on.So my opinion is to limit at firewall filter to 1/32 connections for the specific packet_mark old.At this point i see another problem, at firwall filter we can drop only syn connections with packet_mark old, what about those with estabilished type, cause if i start to download with flashget with configuration of 10 connections, at this moment in connections tab of MT is see 10 connections estabilished. How can i close 9 of these?
P.s. sorry for my bad english. my configuration is excast as descripted in this thread.
Best regards to all from Albania.

31

No one can help at this point?

32

Where have all posts gone??!!

33

Bump.

I’ve been using this, works great, however, does TCP have to be identified in the mangle rules?

I’m having a hard time shaping other traffic separately. If I try to use mangle to prioritize port 80 and give other ports a 256k limit, PCQ just stops.

In order to shape P2P, I have to use another RB. I’ve been doing it on the client side, works well as I only have two problem clients, but I’d like to have a blanket rule so some new customer that discovers P2P won’t go completely nuts on the network.

edit: tinkering with my 2nd AP in line, first doing this as the primary, it appears that choosing the TCP protocol is not required. Any input from those more experienced would be greatly appreciated.

34

What of for hotspot users.

I have 3 interfaces,
ether1 = wan- isp
ether2 = hotspot ip range: 172.16.0.0/24
ether3 = local users = ip range : 192.168.20.0/24

How do i go about it

35

I have the same problem, i make a lot of rules to mark, ssh,p2p, pop, etc. and make a queue for prioritize it, then mark it in postrouting and prerouting, works perfect, but when i setup this or the Janis mangle example from the mum that marks connections in forward the apply PCQ, the pcq simply doesnt work.. somebody can help us? thanks

36

Maybe you are overwriting forward chain marks with postrouting marks?

37

well, the forward chain marks connection based on src-adress-list but the prerouting chains marks packets based on src and dst ports. alos mark some connection to use the connection-bytes but its marked also by src port 80.. any idea?

38

I was having the same problem as mentioned here until I used our new friend connection-rate :slight_smile:

I set the connection rate limit for each connection to be just over the connection speed of the non burst queue. It seems to be working so far. If multiple downloads are started with a download manager, each individual connection never reaches the speed specified in connection-rate so they stay in the non-burst queue. Its working beautifully so far.

Here is a small preview :slight_smile:

add action=mark-connection chain=forward comment="" connection-bytes=\
    0-2000000 connection-mark=JetWire_Conn_1m connection-rate=2M-100M \
    disabled=no new-connection-mark=JetWire_Conn_1m_Burst passthrough=yes \
    protocol=tcp src-address-list="PPPoE Private 1M Service"