Performance issue with HPN-SSH
This is related to HPN-SSH so called High Performance tweaks of SSH.
https://winscp.net/tracker/17 states the feature as being implemented.
Looking at the linked post on the forum, I can see some tests were made but a long time ago (2011).
I made some new tests last week end with WinSCP build 5.1.5.
Transfers were made between a home built NAS running NAS4Free (x64 build 775) and a desktop computer running Windows 7 x64. These 2 computers are connected directly to each other via gigabit LAN.
The test consists of uploading files to the NAS and measure speed achieved with WinSCP on one hand and Filezilla on the other. Using SFTP with AES encryption in both cases.
Here are the transfer graphs generated by Nas4Free.
With WinSCP:
With FileZilla:
As you can see there is gap greater than 6 times in transfer speeds between the 2 software !
Also worth mentioning is the fact that during transfer, WinSCP consumes one CPU core fully (on desktop computer) and constantly.
FileZilla is also showing almost full core utilization but there are variations.
The speeds achieved with Filezilla prove that the NAS is able to receive data at least at that speed.
So it seems there is bottleneck somewhere within WinSCP preventing it from reaching higher transfer rates.
Would it be possible to investigate on this limitation ?
https://winscp.net/tracker/17 states the feature as being implemented.
Looking at the linked post on the forum, I can see some tests were made but a long time ago (2011).
I made some new tests last week end with WinSCP build 5.1.5.
Transfers were made between a home built NAS running NAS4Free (x64 build 775) and a desktop computer running Windows 7 x64. These 2 computers are connected directly to each other via gigabit LAN.
The test consists of uploading files to the NAS and measure speed achieved with WinSCP on one hand and Filezilla on the other. Using SFTP with AES encryption in both cases.
Here are the transfer graphs generated by Nas4Free.
With WinSCP:
With FileZilla:
As you can see there is gap greater than 6 times in transfer speeds between the 2 software !
Also worth mentioning is the fact that during transfer, WinSCP consumes one CPU core fully (on desktop computer) and constantly.
FileZilla is also showing almost full core utilization but there are variations.
The speeds achieved with Filezilla prove that the NAS is able to receive data at least at that speed.
So it seems there is bottleneck somewhere within WinSCP preventing it from reaching higher transfer rates.
Would it be possible to investigate on this limitation ?
