HTB strange results

Experimenting to set up QoS to down prioritize FTP traffic I am geeting some very unexpected results. Despite setting different priorities / setting max limits very low I am still getting the same download speed through port 21 and port 80.

I tried with a lot of different values and is able to limit the traffic to a specified max limit, but not achive that port 80 has a higher priority when max limits are not achieved.

All mangling is done in the prerouting and seems to work fine.

Please let me know if you can see in enclsoed jpg why the download speed is the same for both port 80 and 21.

rgs Pilgrim

max-limit=2?..

2 bits per second?..

max-limit (integer/integer) - data rate which can be reached if there is enough bandwidth available, in form of in/out (target upload/download)

Thanks for the comments I really hope you can help me to a better understanding of this problem.

I made an update to figure with the missing rates.

The problem: What seems to happen is that both Queue03 and Queue04 is getting 1M each despite the different priority setting. Is that an expected result based on the settings shown??

I think the mangle is done correctly as I am able lower the download speed by reducing the max-limit of increasing the limit-at for the other queue.

One more question: if the max download speeed from the ISP was 4M instead of the 2M, would the Max-Limit on queue01 then limit the download to 2M. Or does the max-limit simply means that when exceeded than the queue can’t borrow from it’s pareant?

Sincerely

Pilgrim
HTB test.jpg

read about limit-at parameter.

regards

They way I understand the “Limit-at” is, that it is the “committed rate” and the examples shown in http://wiki.mikrotik.com/wiki/HTB leads me to the understanding that Queue04 will get the committed 128K and Queue03 would then get the rest up to the available 2M because of the higher priority i.e Queue03 would get app. 1.72M.

However, the test we did gave a completely different result. Is above assumption correct? or what should be the expected outcome?

Thanks,

Pilgrim

PS. I know that the way this queue tree is set up is not entirely logical. It came out this way because the first attempt didn’t gave the desired result and we hten started messing around with the parameters to test different settings, until we gave up a bit and went back to the study table :slight_smile:. All comments that will bring us closer to a better understanding of this problem are more than appreciated.

From what I can see in the picture

  1. Queue3 will get 1300k
    Queue4 will get 128k

  2. Queue3 will get the rest of the traffic (till it reach 2M)

If it isn’t so you need to check that you are really marking the right traffic. In global-in there are upload+download and input+forward together, so you need to separate only necessary traffic.

Thanks, I need to check my mangling again then.

To fully understand how the max-limit work - What would happen if:

  • The ISP increases the avaiable speed from 2M to 4M
  • All settings in the queue tree remains unchanged
  • and there is no traffic on Queue03

Clear that Queue02 will not be able to borrow from its parent after reaching the 2M max-limit, but since there is no other traffic will Queue04 then get all 4M - or will the end effect be that max-limit work as a bottleneck and only allow 2M for Queue04?

Thanks,

Pilgrim

2M only

Thanks,

I guess the only thing that can owerwrite the max-limit is if the sum of the child limit-at’s is higher than the parents max-limit.

I was kind of hoping that the max-limit would not be a hard limit and just mean that the queue would not be able to borrow from it’s parent.

Back to the study board again :slight_smile:

rgs Pilgrim