This message was deleted.

More iops to the disks that back longhorn should help.

No, it doesn't. I run speed tests and a local folder on the worker has 1000 MB/s while a folder on the longhorn storage has 100 MB/s

It's also replicating data over the overlay to multiple nodes. There are quite a few factors contributing to performance.

What are the steps that you run to measure the performance? I have done this test on EC2 instances: https://github.com/longhorn/longhorn/issues/2986#issuecomment-923404663