We have database backups that we sometimes have to inject into the databases concerned. If we copy the backup to the VM hosting mysql, then we perform the import, it is very fast (3 minutes on average).
But if we import from the VM that hosts the backups to the one that hosts the mysql, the import is much longer (15 min on average).
The import uses the private network to which the VMs are connected. What is strange is that these 2 VMs are on the same VMware host. The write speed should only depend on the speed of the host disk (10 Gb / s).
So there is something blocking, but we can't seem to find what.