I was told to use PCQ with address lists to limit my PPPOE customers. But lately I am hearing I should be using simple queue with the improvements made in ROS v6.
Which should I use? Continue with PCQ or switch to simple queue? Are there situations where one is better than the other?
I’ve had hundreds of users using simple queues on an RB1100AHx2 on both RouterOS 5.x & 6.x without any trouble. Unless your pushing the limits of your hardware then do what ever works best for you and don’t worry about what is the best performance wise because you’ll be spending your time splitting hairs. Note: This recommendation comes from a perfectionist… so saying not to worry about the best in an area is kind of hard to do. I would spend my time and energy in other areas. For example prioritizing VoIP traffic and network essential traffic like OSPF signaling etc. We have over 3000 customers using PPPoE with Simple Queues on a mixture routers ranging from RB2011 to RB1100AHx2’s with ROS 5.x & 6.x and I don’t regret simple queues at all.
Our billing system assigns the mikrotik-rate-limit radius parameter (which creates the simple queue). It is simple and effective. I don’t really care for my use if there is a more efficient means because this works really really well and has worked perfectly since I set it up over 3 years ago.
The router in the picture below has been servicing over 200 customers using simple queues for about a year without even rebooting.
Thanks for the info. We are migrating to a billing system integrated with freeradius, and I was feeling simple queue was the way to go. The 1100AHx2 is what we are using except in a couple sites that we will be replacing with 1100AHx2.
edit: Also how much bandwidth are you pushing on your 1100AHx2?
The router I pictured was a local router so that router only passes a bit over 40Mbps (7 day chart attached). I have an RB1100AHx2 right next to it that does distribution traffic peaking up to 300Mbps but that router isn’t doing PPPoE or Simple Queues. It is however using the Queue Tree for QoS for backhaul traffic.
Thanks for the help. I started running simple queu’s on the pppoe servers and it’s working well. Except on the RB2011’s I get more throughput on PCQ address lists on the one site that has > 40 clients. But both of those are being replaced with better hardware.
I am considering going CCR1009 instead of 1100AHx2’s for future deployments. Any opinions on the question at hand.
I don’t have much experience with the CCR but I expect that it will work very well for you. The PPC chip is much better then the MIPSBE so don’t underestimate the 1100AHx2 based upon seeing a performance wall with the 2011. Even if the 1100 wasn’t dual core and wasn’t much faster per core just the chipset change would render better performance. I am guessing that you are pushing the 2011 about as far as I trust them with 40Mbps along with queuing etc. The 2011 is more of an endpoint firewall then a infrastructure plus qos & PPPoE concentration. I would recommend using the 2011 to build small sites but always plan on replacing them once growth warrants it.
This is one of my gateway, this one has 100Mbps full duplex.
Notice: I reboot automatically the machine each 28 days @ 04:00 AM
I use symmetric profile for user with more bandwidth compared to other picture…
I do not know why some WISP provide asymmetric ADSL style profile, on actual wireless technology.
(Until TDMA GPS sync tecnology can be used easily…)
dear joshaven can u post simple queue screenshot details and some screenshots for voip qos!
For VoIP (we run our own VoIP server) I match any traffic coming or going to our VoIP IP range. I split this traffic into VoIP management and VoIP traffic. I also match any network management traffic and UDP or TCP bandwidth tests and any traffic with an existing DSCP bit. I then use the tree queue to outbound traffic on the backhaul interface or AP interface setting the maximum sustainable speed as the max limit. Under the interface I have a queue to match up with each packet mark and I give limit-at values to VoIP traffic as well as network management. The rest is managed by priority.
The setup is very particular to the exact situation so I don’t think that screenshots will be that helpful. Feel free to ask questions. I’d also be willing to skype sometime regarding your setup and I could give you a recommendation specific to your network.
Furthermore the QoS will not be applicable to traffic inside of the PPPoE tunnel because the traffic is aggregated. To solve this we setup a VLAN on the equipment and installed two internet connections for our VoIP customers… the PPPoE tunnel (untagged) would be standard traffic and would be delivered to the CPE with a public IP. The VoIP (tagged traffic) was all on network so we delivered this with private IP’s that were not using NAT. This solved all of our issues regarding call prioritization. It didn’t stop the clients from having bad connections so that has to be always maintained but that is nothing new to a WISP…
For simple queues… I use radius with the Mikrotik-Rate-Limit param which is handed out by freeradius. I am also pulling the user info from our billing system.
Usage info here: http://wiki.mikrotik.com/wiki/Manual:RADIUS_Client
Look for: Mikrotik-Rate-Limit