Reduced VLAN performance on WAN port of MOX

MOX hardware MOX HW problems
1

My ISP has glassfibre with a passive optical network so I have to connect to its copper terminal, which exposes Internet on VLAN 6 (I think TV & telephony are on different VLANs but I don’t use that).

So I was able to use LUCI to use PPPoE on the WAN port and use eth0.6 as the interface instead of eth0, which works fine, I get Internet.

But now I have the luxury problem that my Internet caps out at 450 mbit/s using the MOX while with the ISP’s route I get up to 930 mbit/s. At 450 mbit/s the MOX is already at 100% CPU.

I suspect the VLAN must be the problem because if I put my ISP’s router in between I do get better performance (not the full 930 mbit/s but closer to it).

Does anybody know any tricks for getting higher performance out of VLAN stuff on the MOX? Would I be better served with an Omnia (because it has a higher CPU frequency)?

2

I did tests on the MOX that show there is no performance degradation purely because of VLAN-traffic without NAT: MOX wired throughput issues

NAT-traffic automatically reduces this possible traffic because of involving the firewall.
The figures you share seem reasonable to me. A Turris Omnia for sure can reach full bidirectional gigabit NAT ethernet throughput (without traffic shaping, but I believe that’s nothing you after thinking as of now, right?).

3

PPPoE surely plays a role there as well

4

Right so it’s probably the firewall or PPPoE or both?

As far as traffic shaping goes, I have left that on the stock settings of TurrisOS - I’m not exactly a networking expert (yet? :D)

I vaguely recall seeing that tc qdisc controls traffic shaping so here’s what it’s set to for me:

root@turris:~# tc qdisc show dev eth0
qdisc mq 0: root
qdisc fq_codel 0: parent :8 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
qdisc fq_codel 0: parent :7 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
qdisc fq_codel 0: parent :6 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
qdisc fq_codel 0: parent :5 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
qdisc fq_codel 0: parent :4 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
qdisc fq_codel 0: parent :3 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
qdisc fq_codel 0: parent :2 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
qdisc fq_codel 0: parent :1 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn`

I did try disabling Sentinel (normally I’d prefer to have that on) but it didn’t really change anything.

Maybe I’ll just wait for Omnia 2022 and buy that then, especially if it also has wifi6 (I did all my speed tests over wired ethernet to be clear).