Speedtest is what a customer make to verify his profile, and this test open 4 simultaneous tcp connections.
So when I try to emulate from a RB to a CCR using Mikrotik BWTest TCP with 4 sessions I obtain the exactly result
Now I ask if you experience the same behaviour, and why there is this very, very big different value between Cambium test and real performance of the radio link
AP is with only 3 SM connected and make no traffic when I make the test. All have firmware 2.6.
And RB and CCR 6.35.2 (no cpu to 100%)
There is something I can tweak to optimize single user experience?
I'm sure you have a lot of retransmission on uplink. Retransmissions in uplink impacts your download because of TCP traffic nature that requires ACK.
You clearly have interferences on the AP on the frequency of your AP and not on SM side (because your download speed is high).
Please do a spectral scan to check this on both side.
If you can't do better because it's the "best" frequency possible, I suggest you to look at the "Performances" section on the SM in Uplink, and look at the MCS that has the highest percentage of traffic.
Set your SM Uplink to that MCS as a fixed value: your TCP traffic will go up and your link will become more stable.
I noticed that the MCS selection algorithm doesn't work so well in high interferences environment and it's better to manually set uplink to have a good result in TCP speed tests on speedtest.net.
I also see your signal is very high (-44), so maybe you have saturation due to high signal. If your spectrum is clean, try to lower the power or to enable auto power on the AP.
If you don't have frequency reuse enabled and you're using auto power, you can set your target signal to a higher value than the default -60dBm to have a higher difference between noise and signal.
You can also use eDetect to show co-channel interferences on both sides of the link.
I'm fighting this issue right now at two different tower sites. At both tower locations I can run a speed test and get ~50 down and 70 up. At the tower I am running EPMP 1000 GPS with frequency reuse and 50/50 ratio. Customer location we are currently placing Force 200's but I'm pretty sure I have seen this on a Force 110 as well. On a site we turned up today I get 25/20mb on a Wireless Link test. Speedtest.net I get 8/20mb.
I'm fighting this issue right now at two different tower sites. At both tower locations I can run a speed test and get ~50 down and 70 up. At the tower I am running EPMP 1000 GPS with frequency reuse and 50/50 ratio. Customer location we are currently placing Force 200's but I'm pretty sure I have seen this on a Force 110 as well. On a site we turned up today I get 25/20mb on a Wireless Link test. Speedtest.net I get 8/20mb.
are you using end to end flow control in your switching/routing system?
I've commonly seen that be the underlying cause.
you'll also beable to observe it from the local Ethernet switch to the AP, if you've got systemical bandwidth, do a speedtest, on a 100 meg port, you'll see something like 50 or 60 meg down, and 90 up. if that's the case, check your flow control and test again, you should see a decent result.
there are other possibilities, that's just the most common reason from my experince.
Yeah, I put in a Netonix switch at my main tower that has a mixture of FastE / GigE devices for this specific reason and flow control is working. My other tower sites are 100% GigE.
The path from the nearest router all the way to these radios as well as any device carrying large amount of data within that l2 domain to get its full benefit. One someone what choking device can be detrimental to everyone’s performance.
Plugging I to your switch next to the ap with 100 meg interface will show backward speeds as well if FC is your issue