I’m hoping that somebody will be able to assist me with this. I have reached out to MikroTik support as well but they seem unwilling to assist.
I have Two point to point SFP+ links between 3 buildings. Building 1-2 and Building2-3. Each building has a CRS with a 10G-SFP31-LRM-C1 in the SFP slot to convert the fibre.
I have tried with auto-negotiating off and by specifying 10G speeds.
I have also let it auto-negotiate and for both links it chose 10G.
I am only seeing speeds of 500Mbps on these point to point links where I’m expecting much more. I have put input and output chain accept rules at the very top of the firewall to bypass any firewall implications but that has not helped. And the bridge setup is very simple so I’m not expecting that to be an issue.
Please find attached a screenshot of the detected SFP parameters.
How do you test?
What exact CRS do you use?
What is the CPU usage while testing?
What RouterOS version are you running?
Can you share your config (/export hide-sensitive file=anythingyoulike)?
I used bandwidth test with UDP settings from the central CRS that connects to both other buildings. I am running receive as a separate test and then send as a separate test. Default bandwidth test params otherwise.
All three are CRS326-24G-2S+.
Running long-term-stable ROS - 6.45.9. A new one has just launched midway through my testing, but I did not want to update just yet as I'd have to restart all my tests.
CPU usage on the CRS that I'm running from does go very high (88%) so that's likely a restriction. Suggestions on how to reduce the cpu usage?
I've run out of time for today but I'll post the config tomorrow.
Simple solution:
Do not use the devices for testing (well, better: guessing) bandwidths. It’s always a CPU bottleneck.
Get yourself two computers, connect them to the switches and run iPerf on them.
-Chris
Writing to support about improper use of equipment is pointless.
Those CRSs are switches, not routers.
Any firewall rule cripples them. Check https://mikrotik.com/product/CRS326-24G-2SplusRM#fndtn-testresults
Bandwidth tests ran from devices with a single core CPU running at 800MHz is also bad. There is no way those can generate 10Gbps of traffic.
Fire whoever designed that network and don’t let them touch network equipment ever again.