Hello, we are still encountering the same issue even after updating to the latest stable version. The configuration in RB is the all-in-one type, where PPPoE Server, NAT, and Simple Queues are in the same box. The processes are still stuck on the CPU0 and Simple Queue is not multi-threading, even though it was designed to multi-thread. Doing a bandwidth test on 127.0.0.1 seems to utilize all the quad cores. I've also reported the issue to the support, stating it could possibly be a bug, but it seems they can't find anything useful. Is there anything else to do with this? Why is this only happening at ARM, but not with other kinds of architecture? IPoE connections works fine, but not when using PPPoE for some reason. Does anyone in v7 still find the same issue as well?

Currently, we are thinking of replacing our RB with CCR1009, 1016 or 1036 because we are stuck with the throughput of only 400Mbps on average.

as far as i know simple queue is NOT multi-threaded

this problem arise frequently when a single PPP interface plus another config when most traffic pass across that PPP interface

as far as i know simple queue is NOT multi-threaded

this problem arise frequently when a single PPP interface plus another config when most traffic pass across that PPP interface

Any alternative solution? Should we buy another RB and separate the PPPoE Server? Or just get CCR and avoid ARM based CPU in all-in-one type?

RB's ether-1 and 2 has PPPoE authenticated from the ISP, then I've only bridged the remaining ports and placed the PPPoE server on the bridge itself. Before we migrated to the authenticated PPPoE from the ISP, we ran the traditional DHCP on the given modem/router from the ISP. The CPU's load was okay, but they were not evenly distributed, most process was still bound to CPU0. It became worse when we transitioned to 2 WAN PPPoE, where all the processes are stuck to CPU0. I'm disappointed with the performance of RB in this type of configuration.

as far as i know simple queue is NOT multi-threaded

Yes, is it, at least on other CCR devices I have on 6.46.8 and 6.48.6

I must be specific

Each simple queue or a single parent queue is executed by a single core

If You have your load distributed across multiple queues with different parents You can expect to be multi threaded

Beyond this in this specific case the problem arise mostly because of the ppp interface