Really slow zfs send & receive withUbuntu 22.04

Hi guys, is there any known issues I need to be aware of when doing a zfs replication over ssh between two Ubuntu 22.04 servers?
The speeds are down around 170k per step over ssh. The hardware on both servers is quite good and not reporting any load to speak of. The network reporrts about 21Gbps over iperf3 and if I do a scp from the cli it’s running 550MB/s to the other box.
If I dump the zfs send out through pv to ‘dev/null it seems to concur nothing is wrong on the sending side.

Could it just be ssh as the problem in this use case? I’ve not had this problem before.

You might be hitting issues with TCP window scaling. You didn’t say how you’re managing your replication, but if you don’t have any network buffering in the loop, adding some might go a long way toward mitigating your issue. (This is why syncoid uses mbuffer.)

You could also try HPN-SSH, especially if there’s a lot of geographic distance (or other reason for high latency, despite high throughput) between your servers.