Question Iperf3 results: Significantly different speeds uploading vs downloading?

[EvanVanVan]

I'm trying to track down slow streaming speeds from my Plex server to my parents house. Both the server and client are technically on a 1Gpbs connections.

The server is FiOS and speedtests routinely return (940/380) speeds (it really should be 1gbps upload too but it's suprisingly slow in my current apartment.)

The client is Comcast and shows (900/40), which is the expected upload speed.

Testing with iperf, the client to server pretty much maxes that 40mbps upload:

But in the other direction, server to client, transfers are very slow:

Any suggestions? I guess the first thing would be talk to Verizon and see why my ul isn't reaching 1gbps...

Thank you

 

[bill001g]

So it might not matter it depends on what you call client and server and how that relates to what exact options you used in iperf.

Iperf has given me strange results in the past. It been a while since I used it but I think it defaults to download traffic from the server to the client. This is based on how you start IPERF not what you are calling the machine.

What I have seen is you get different results if you were to just start one machine in server mode and then test a upload and download compared to swapping which machine is the server and then always doing download.

Now maybe it doesn't do this anymore but I never 100% trusted the results.

How do you connect the 2 location do you have a vpn or are you using something like port forwarding to get it to work. Either of those could also be to blame.

Again its been a while but there used to be a internet based IPERF site you could test from. It would be interesting to see what each of your machine test to that site. It would to a point rule out IPERF.

In general if speedtest gets 380mbps you other software should get that also. Version will blame software which to a point is valid. It does not though explain why you do not see more than 380mbps. Make sure you run speedtest on a number of different sites best if you can find one run by verizon. I know when I was experimenting on a friend who paid the big money to ATT for a 5g fiber you could only get those results if you tested to ATT everything else barely was above 1gbit.

 

[EvanVanVan]

So it might not matter it depends on what you call client and server and how that relates to what exact options you used in iperf.

Iperf has given me strange results in the past. It been a while since I used it but I think it defaults to download traffic from the server to the client. This is based on how you start IPERF not what you are calling the machine.

What I have seen is you get different results if you were to just start one machine in server mode and then test a upload and download compared to swapping which machine is the server and then always doing download.

Now maybe it doesn't do this anymore but I never 100% trusted the results.

How do you connect the 2 location do you have a vpn or are you using something like port forwarding to get it to work. Either of those could also be to blame.

Again its been a while but there used to be a internet based IPERF site you could test from. It would be interesting to see what each of your machine test to that site. It would to a point rule out IPERF.

In general if speedtest gets 380mbps you other software should get that also. Version will blame software which to a point is valid. It does not though explain why you do not see more than 380mbps. Make sure you run speedtest on a number of different sites best if you can find one run by verizon. I know when I was experimenting on a friend who paid the big money to ATT for a 5g fiber you could only get those results if you tested to ATT everything else barely was above 1gbit.

Yeah, I'm not 100% certain which direction the transfer is going I think I have it right with the iperf client transfering to the iperf server (client and server being the same as plex client and servrer respectively). But the I also did the test with -reverse so it should be going in both directions, regardless.

Currently, I opened a port for the test. It wasn't through a wireguard tunnel.

I'll look into other options to confim my results though, ty.