I see you are just using the default exactly pretty much. But this switch is connected to a router, yes? What about the router config?
To test the switch properly you should be doing testing without going through a router - do a speed test between two computers going through the switch with iperf3. That way you aren’t possibly introducing other unrelated problems that are botching your test.
6.45.7 is new, and I would generally suggest people use the long-term release in most cases instead of the “stable” for production environments.
Also I hope that you are updating your firmware too, not only the RouterOS version. If you are only updating RouterOS you might be behind on the firmware updates.
I’m happy to perform any test to continue troubleshooting these issues. I have done point to point file transfers before with similar results some ports perform well with high throughput and low ping and others with low through put and high loss
So first thing, make sure your firmware is upgraded, not only the RouterOS version, otherwise you may be troubleshooting phantom bugs that should have been fixed.
Second if everything is updated is to unplug the switch from everything and do port to port speed tests with only two computers plugged in on 10Gbps links, using iperf3.
All MiktoTik equipment we are using has v6.45.7 as the OS and the Boards firmware updated and we tested with 6.44.6 and 7.0B3 all came back with the same results and same issues. Also all equipment was factor reset with minimum changes to make the network usable.
That said tomorrow we will test with iperf3 and report our finding.
OK. One other question, how is the switch connected to the router? What port on the router is going to what port on the switch? Are you using a DAC cable for that or something else and if so, what?
CCRs is connected via the SFP + port connected to the CRS SFP + port 1 as configured as a Lan bridge. And all their Network equipment connects off of the 10-gig switch through SFP + ports. We have multiple aoc’s and dacs test results don’t vary depending on what we use both come back the same results. Normal scenario we use all AOC or all DAC and don’t mix and match on the network not that should have any impact. I have not tried running the tests off of a SFP+ to 10gb ethernet yet as the test didn’t change between AOC and the DAC I didn’t bother SFP+ to ether net.
OK, was just making sure that you were not trying to plug into one of the regular SFP (instead of SFP+) ports on the CCR.
You’ll have to do the iperf3 tests on the switch by itself. I use this switch model at the core of an ISP doing about 1Gbps traffic and do not have loss or any performance issues.
Found the issues to be the use of a MikroTik compatible DAC cable that was not compatible with the NIC. So I switched out the SFP Transceivers. Issue solved
This issue could be caused by the CRS317 being picky about what it connects to. When connected to the CRS328 my CRS317 the link is flapping with lots of FCS and RX errors. Are the links to your CRS328 stable (check the link down counter for the interface to the CRS328)?
Are the Aquantia’s stable when you connect them to the CRS328 SFP+ ports?
