Issues with Simple Queues

Hi, I’ve set up a bridged RouterOS computer, one eth in and one out.

I enabled IP-firewall to get the simple queues working, not using packet-marks, just using the simple queues with their target-addresses.

We ended up creating one final queue for all addresses that isn’t in one of the other queues, this seemed to be working fine when testing with 10-15 IPs, but is acting up quite badly now when running it for real.

basically we have 400+ queues like:
name=“2527-01-307” target-addresses=10.100.90.10/32
dst-address=0.0.0.0/0 interface=all parent=none direction=both
priority=8 queue=default-small/default-small limit-at=0/0
max-limit=4M/8M burst-limit=0/0 burst-threshold=0/0 burst-time=0s/0s
total-queue=default

The final rule is for 10.0.0.0/8 and is limited to just 20kbit/s.

This is where things go wrong, connections seems to be randomly limited to 20kbit. Some are working just fine, some are working for a while and then gets dropped from their real queue-speed to crawling 20kbit.

I figured the Queues worked like the firewall rules and was checked from lowest # to highest and first Queue that matches gets it, but now it seems more like it picks a queue at random that matches and considering the 10.0.0.0/8 rule always matches people sometimes gets put in there.

The reason for going with just the simple rules instead of using packet-marks was for simplicity, I was going to make a webpage for adding IPs/Queues using the API and using packet-marks would make that a bit harder.

The question I’d like to ask is, should this work or will I have to change to packet-marks instead?

We’re using Routeros-x86 v4.9

Regards
/Jonas

I know this isn’t a direct answer, but do all those 400 queues share characteristics (do they all have the same rate applied, or are there 3 or 4 different rates that are applied to lots of users)? 400+ simple queues aren’t going to work all that great, but if the queues are similar then PCQ would be fairly easy to deploy and will work significantly better.

there are a couple of different rates.. but it’s 3 that are the most common.. then only 10 users or so have different rates..

we started with pcq, the main reason we skipped those was that some of the users have multiple IPs, and we wanted all of those share the same rate.. with pcq each IP would get the same speed so a user with 3 ips would get 3 times the bandwidth..

thing is, i dont think it has anything to do with there being too many queues, or the router not being able to handle it.. since it’s working perfectly well if i disable that last 10.0.0.0/8 rule.. the only problem then is ofcourse that someone using a nonlisted IP will get full speed