IPerf3 slow to zt IP and fast to real IP

Im testing hosts at two locations, 80 km apart on different ISPs.

Iperf3 gives me speeds close to the wire 1 Gbps limit when using the hosts zt address:

Accepted connection from 192.168.196.122, port 16871
[  5] local 192.168.196.241 port 5201 connected to 192.168.196.122 port 16872
[ ID] Interval           Transfer     Bandwidth
[  5]   0.00-1.01   sec  11.0 MBytes  91.7 Mbits/sec
[  5]   1.01-2.00   sec  11.0 MBytes  92.6 Mbits/sec
[  5]   2.00-3.00   sec  11.2 MBytes  94.2 Mbits/sec
[  5]   3.00-4.00   sec  11.2 MBytes  93.8 Mbits/sec

But only 100 Mbps when using the physical network address:

Accepted connection from 192.168.196.122, port 17415
[  5] local 192.168.1.207 port 5201 connected to 192.168.196.122 port 17418
[ ID] Interval           Transfer     Bandwidth
[  5]   0.00-1.00   sec  1.05 MBytes  8.82 Mbits/sec
[  5]   1.00-2.00   sec  1.12 MBytes  9.40 Mbits/sec
[  5]   2.00-3.00   sec  1.12 MBytes  9.41 Mbits/sec
[  5]   3.00-4.00   sec  1.14 MBytes  9.61 Mbits/sec

When listing peers i see no relays, all connections are direct.

The virtual adapter on the iperf3 server is saying its running at 100 Mbps, but i read in the official docs this doesnt reflect its actual speed. Not sure whether the above test shows this is true or not or what…

I should mention that the PCs im testing with are running Windows.

If they are really directly connecting, the other bottleneck is CPU. Is ZeroTier using 100% of a CPU on either machine?

1 Like

Is there some more real way to understand if they are directly connecting beyond the cli peers commands?

The original test was the server on a Celeron J1900. Thats 4 cores, 8 Gb RAM, theres negligible cpu use according to windows task manager.

There are from a Ryzen 5 5560U, 6 cores, 12 threads, 16 Gb of memory on the same subnet:

Connecting to host 192.168.1.149, port 5201
[  4] local 192.168.196.116 port 10727 connected to 192.168.1.149 port 5201
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-1.01   sec  1.00 MBytes  8.34 Mbits/sec
[  4]   1.01-2.01   sec  1.25 MBytes  10.4 Mbits/sec
[  4]   2.01-3.01   sec  1.25 MBytes  10.5 Mbits/sec
[  4]   3.01-4.01   sec  1.12 MBytes  9.42 Mbits/sec

And with the zt address:

Connecting to host 192.168.196.129, port 5201
[  4] local 192.168.196.116 port 10766 connected to 192.168.196.129 port 5201
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-1.00   sec   896 KBytes  7.34 Mbits/sec
[  4]   1.00-2.01   sec   384 KBytes  3.10 Mbits/sec
[  4]   2.01-3.01   sec   512 KBytes  4.21 Mbits/sec
[  4]   3.01-4.01   sec   384 KBytes  3.15 Mbits/sec

So for some reason it is now all slow… but the Zerotier ip is still noticeably slower.

I think the peers list saying “direct” is the best way to tell. The connection would probably be even flakier than that if it was relaying.

The Celeron doesn’t have AES hardware acceleration, so it’ll be using CPU to do encryption.

It’s hard to guess what’s going wrong! Might be something in the physical network. Is it peering through the internet router, instead of directly over the LAN?

1 Like

Im not sure… how do i tell?

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.