One link (dual mimo antennas, 3km, on each side 20 networks), throghoutput 80Mbit/s.
Second link(one polarized antennas, 2km, on each side 20 networks)
check the ethernet connections between routerboards, if the link speed is 1Gbps between Routerboard 1 and 2 , and 100Mbps between 2 and 3 then you could have this problem.
Just to try : force all ethernet connections to 100Mbps FD
I fight this problem for a long time now. It is not ethernet in my case. It is nv2 and queuing. Nstreme was better but do no longer work good with newer versions.
I tried every queing method and different queue sizes. May be it is how small tcp acks are grouped in nv2 packets.
It adds up with every hop and it is different in up and downlink.
No. As I have newer ROS installed nstreme does not work well. I see disconnects. The same
hardware had no problem with nstreme and an older version of ROS. I see good speeds with
nv2 and UDP bandwidth tests. So the capacity and reliability of the links are better now but the
tcp-bandwidth is way below what would be possible.
The problem still persists, I tryed v6.17, v6.18. If traffic goes over more than one router(connected with ethernet cable) throughoutput is abou half…also tryed turn off auto negotiation and fixed 100FD…without success…
Same situation with 6.15 on some links. Some links are a bit better.
I’ve ordered some more SAF Links. It feels sane when you install a link
and know exactly what will happen and if something goes wrong you know
it is your fault.
Ordered a test link of a quantenna based equipment.