I have several PPPoE servers (RB800, v5.20), and installed a new RB1100AHx2 (v5.20).
I tried migrate 3 of RB800s (about 100 users per board) to one RB1100AHx2, but I have performance problem.
Queues (per pppoe user) that have ‘high’ bandwith settings (between 4 and 20 Mbit/s) are very slow. If i disable the user queue, then no performance problem, user can download at full speed.
The default pppoe queue type is ‘default-small’, but this is not enough, I set to SFQ (preturb 5s, allot 1514) it’s working for about 120-150 users but no more.
Is it possible that the RB1100AHx2 can’t handle 250+ users (CPU load 50-60%, firewall tracking is off) or settings problem on queues?
Any idea?
hello
RB1100AHx2 have problem in multi cpu supporting.you can upgrade your version to 5.21 and test again.or wait to V6.X brings in mikrotik website.
also you can change queue type to multi-queue-ethernet-default.
This sounds like tragicomic. Is the ‘top level’ router as the worst choice? LOL.
I can buy six or seven RB2011 routers and get more performance at equal cost… Brrr..
I’m very curious about the performance of the new CCR router.
The solution that macgaiver proposes is the best approach. In pppoe server you have the option to add the clients on to address lists, and from there you can build queue tree, which is far better than simple queues. That was the idea behind the option of adding clients to address list. Also, you should not specify any limits on the pppoe-client, so the simple queues are not automatically generated.
I was having the same issue on RB1100AHx2 routerOS v5.17 with around 250 customers and ~120M load during peak hours.
We are using simple queues; radius policing with the following attribute: Mikrotik-Rate-Limit = “6000k/15000k 6144k/15360k 3072k/7680k 10/10”
Edit: We do have a wide variety of rate limits.
We have upgraded our RB1100AHx2 to 6.0rc11 7w3d ago. Everything is working perfect.
There are over 250 simultaneous PPPoE sessions generating over 100Mbps of traffic at the moment.
