Odd throughput issue: Link capable of 27Mbps, passing inconsistent 4-8Mbps down 4-20Mbps up

Have been having an issue with a customer on a dedicated PTP 650 link.  At my whits end as to what the heck is wrong. Basically the speedtests to the customer are far below what the link is capable. Seeing 4-8Mbps down and 4-20Mbps up. Every test is diffrent and no test shows full speed. I have confirmed that bandwidth is making it to the Master side of the link, but it cannot get beyond that. Any Ideas? I am at a loss.

Thanks for the enquiry.

I'm assuming that this link is successfully installed and aligned with a line of sight path, as shown in the LINKPlanner project. The Link Loss (122 dB) reported by the PTP 650 doesn't quite make sense. A line of sight link at 5.8 GHz and 11.2 miles has a loss of 133 dB. It's possible that you have used larger antennas, and not updated the antenna gain on the PTP 650s. Could you check?

The ODU is transmitting and receiving at the top modulation mode of 256QAM 0.81 Dual. The reported capacity is correct for this modulation mode, 20 MHz bandwidth and the Lite license.

Could you tell us some more details about how you are measuring end-to-end capacity?

Thanks, Mark

The link probably does have larger dishes on it "3rd party install, but I am stuck managing it" Unfortunately I can't get the gain on those dishes until later this spring due to site restrictions "snow".

Testing was preformed with a direct attached laptop. I ran Speedtest.net tests, DSLreports test, Speakeasy/megapath, the whole shabang. Come to think of it I also ran Iperf3 testing.

Here is some oddity, When I initially started testing I was set to a 172. address that I have NATed at my edge. The initial testing showed full capacity of 13x13 through iPerf and about 12x12 in speedtest.net "at this point the link was set at a  10 MHz channel". I then tried changing the IP on the laptop to the public address  assinged to the customer router. Speeds at that point dropped down to around 8mbps and became inconsistent. After confirming that traffic limitation was setup exactly the same for the public IP as it was for the 172 I tried testing through the customer router using the 172. address on the router. Speeds became inconsistent on the 172. SO tried switching back to the laptop, at which point the throughput continued to suffer as it had through the router. At this point I expanded the channelwidth to 20mhz and repeated the testing. End of troubleshooting for that night.

Come morning time I made  contact with the customer who notified me that randomly through the night they were able to uplodad at 21Mbps  but the download was still hovering in the 4-8mbps range and the upload was hovering in the 10mbps range with spikes and dips at random intervals.

I know it doesn't quite sound like the issue is on the 650 link, but this is an edge case that is not reflected elsewhere in my network. The only diffrent variable I can come up with is the link. 

Any Ideas or directions I can take troubleshooting?

If I understand correctly, the test tools you mention are used to check the connection from the subscribers premises to a remote server. If that is the case, the problem could occur at various points in the network.

Ideally you would check just the PTP 650 link on its own, with the rest of the network disconnected, using dedicated network test equipment or possibly iPerf set to UDP running on two fast PCs.

If that is impossible to do, would it be possible to confirm the performance using the consumer test tools at the core side of the PTP 650 link? In other words, the same network as before, but excluding the PTP link.

The apparent loss of capacity might be because (a) some other unwanted traffic is taking up capacity in your PTP 650 link, (b) the capacity of the path to the server is restricted by a bottleneck elsewhere in the network, or (c) the capacity is good but packet errors are disrupting speed measurements based on TCP.