Recently built a new Linux server running Ubuntu 7.04; OpenSSH server is 4.3p2, using OpenSSL 0.9.8c.
When copying files from Opus via SSH to this box, I am getting horrible performance - 300k/sec max. However, from the same Windows machine, both WinSCP and Dreamweaver seem to run fine. Also, scp-ing to and from another Linux box to the other in question is fine.
So it seems like Opus is the only one that's being slow. Has anyone else run into this? Going to another server on the same network (running a different version of OpenSSH/SSL) is fine, so I can only take a guess that it's a goofy interaction between Opus and this particular OpenSSH version. ?
Weird. The only things I can think of are that it might be the buffer size that Opus is using causing problems with this particular server, or the version of the SSH/SSL code that Opus is linked against. I don't know a good way to test either theory though.
Just to discount anything silly, you're transfering the same file to the same place in all your tests, right? And you haven't set an Upload Speed Limit for this particular site in Opus, have you? (Speed tab in the FTP bookmarks. Check under both Default Settings and the specific entry for the site in question.)
Yup, same file to test (or even not, I get the same results no matter what it is). No speed limit, I don't even see that tab you mention (is that an Opus 9-ism?) Everything is pretty much default.
At the end of the day it doesn't matter too much, as most files will be transferred to this server via samba, but this behaviour certainly puzzled me.
There should not be any significant difference but uploads are often slower due to bandwidth issues.
Don't be confused between FTP over SSH with SCP over SSH. The FTP protocol imposes some extra processing and is slower than SCP. Opus supports SFTP nor SCP.
If you're using Vista, try running this from an Administrator command prompt:
netsh interface tcp set global autotuninglevel=disabled
I'm not sure if you then need to disconnect/reconnect your network connection or reboot so I guess rebooting afterwards is best to be sure.
This command fixes or speeds up a number of network-related things for some people. The autotuninglevel feature seems to cause problems for a lot of people (myself included), depending on their network hardware and the software being used.
If you find it makes no difference then you can go back to the default settings by running this command:
netsh interface tcp set global autotuninglevel=normal
I've read about it slowing down particular FTP clients, for example. For me, my Windows Live Messenger kept disconnecting after a while but still appeared to be connected (yet nothing worked and nobody could message me). After I changed the setting it works fine again. I don't know for sure if it'll help in your case but it seems to fix so many different network related problems that it's worth a try.
The only things I can think of are that it might be the buffer size that Opus is using causing problems with this particular server, or the version of the SSH/SSL code that Opus is linked against. I don't know a good way to test either theory though.