Hi,
I have an issue with an ftp-server which I want to transfer to proxmox.
The performance of ftp is very good - from standalone pc to VM 47 sec for an 4GB file. But if I try to transfer the same file via ftps (port 990) the performance is very bad (after 7 min. 11.6% transmitted) with several reconnects.
First I think that the distro/ftpd has the problem, but with the same config on a standalone server i get much better values.
As test I use an VM as client directly connected via an dummy bridge on the pve-server and transfer the file in 140 sec without reconnects. With ftp take the transfer 57sec (10 sec more than from a standalone ps - it's ok, because the data must read and write from the same storage). 140sec for much cpu-work is also ok.
Then I tested different bridges: a bridge connected to a 10GB-Interface is much slower ( 0.8% in 7 min.) as an bridge connected to an 1GB-Nic.
With and without vlan-tagging on the 1GB-Nic make only a small impact (11.6% in 7min or 15% in 7min).
As VM I had different kvm-vms (devil-linux and debian-squeeze) tested. Ftpd is allways vsftpd and as ftp-client filezilla.
Due to the good performance with the direct bridge and also with ftp I don't know where the issue is.
Perhaps something with fragmention...
Has anybody an hint?
Will try the same with open-vz (but normaly no real choice for that server).
Udo
I have an issue with an ftp-server which I want to transfer to proxmox.
The performance of ftp is very good - from standalone pc to VM 47 sec for an 4GB file. But if I try to transfer the same file via ftps (port 990) the performance is very bad (after 7 min. 11.6% transmitted) with several reconnects.
First I think that the distro/ftpd has the problem, but with the same config on a standalone server i get much better values.
As test I use an VM as client directly connected via an dummy bridge on the pve-server and transfer the file in 140 sec without reconnects. With ftp take the transfer 57sec (10 sec more than from a standalone ps - it's ok, because the data must read and write from the same storage). 140sec for much cpu-work is also ok.
Then I tested different bridges: a bridge connected to a 10GB-Interface is much slower ( 0.8% in 7 min.) as an bridge connected to an 1GB-Nic.
With and without vlan-tagging on the 1GB-Nic make only a small impact (11.6% in 7min or 15% in 7min).
As VM I had different kvm-vms (devil-linux and debian-squeeze) tested. Ftpd is allways vsftpd and as ftp-client filezilla.
Due to the good performance with the direct bridge and also with ftp I don't know where the issue is.
Perhaps something with fragmention...
Has anybody an hint?
Will try the same with open-vz (but normaly no real choice for that server).
Udo