My ISP offers me 4 gigabit fiber (symmetric). This comes into the house into a box that turns it into a RJ45 port.
This is connected to one of the SFP+ ports on the router through a transceiver. Then there’s another SFP+ port on the router that goes to my main switch.
When I try to do a speedtest from a client, I get about 3900 mbit/s download, but only 500 mbit/s upload, which is supposed to be a lot more.
I have a PPPoE client set up as my ISP requires this.
Not sure what other information I need to provide, please let me know what and I will be happy to provide it.
Update with extra information:
The SFP+ modules used: MikroTik S+RJ10 SFP+10G Copper Ethernet Transceiver
RouterOS version: v7.15.3
Yes it’s supposed to be symmetric.
I did a speedtest and watched the CPU usage. During the download test the highest I see is 26.4, during upload the highest I see is 3.5.
As I was preparing to reply to you my link on the LAN side started to go up and down constantly.
There’s a patch cable that goes from my router to a TP-link switch. The transceiver on the TP-link switch was so hot I could barely touch it to get it out. I think that side may be overheating and another issue I need to troubleshoot first.
Since firewall filters are processed top-to-bottom, this rule has to be above similar rule with action=accept to be effective.
Sadly fast-track for IPv6 doesn’t exist.
Mind that fasttrack rule helps when throughput is limited due to CPU being used to the limit. If your router uses low CPU (e.g. less than 50% average with none of cores peaking to 100%) for moving traffic, then fasttrack won’t improve speeds as the bottleneck is elsewhere (e.g. your initial bottleneck was overheated SFP+ module).
In your particular case you may be near (or at) the limit due to PPPoE requirement. ROS handles PPPoE encapsulation/decapsulation on single core … could be that both directions with same core and in this case full-duplex speed might suffer (as compared to speed when only one direction is heavily used). Fast-track has nothing to do with this issue as fast-track is (mostly) L3 feature
It seems like a fasttrack rule is already in place (the default one, I just noticed it in the firewall rule list) and the packets/bytes counters do go up when running a speedtest.
It’s odd as the download speed results to ~3700 mbit/s, and the upload is ~678 mbit/s which is a significant difference. I was just wondering whether there was anything I could do about it.
I did watch the CPU usuage when doing a speedtest and it’s not like it is hitting 100% on the upload part of the test, it’s around 20 during download and doesn’t even hit 10 during the upload.
If observing CPU usage under system resources … that one is average and single-core tasks won’t trigger it to go very high (depending on number of CPU cored in your device). It’s better to run CPU profiler to see, if one of CPU cores gets pegged and which process causes it.
The results are that without it, the download speed is about 2400 mbit/s. WIth this rule in place, the download speed is 4000 mbit/s.
The upload speed however, is mostly unchanged at around 600 mbit/s, which makes no sense to me.
I also don’t see any high CPU usage, this is a screenshot while doing the upload part of a speedtest (I assume this is what you meant with profiler?):
