Why are SFTP bridges so painfully slow

bridgesftp

Some of our users want to mount folders on our server to a drive letter or mountpoint on their Mac or PC.

SFTP bridges (Trasmit on Mac, WebDrive on Win, etc) are painfully slow to navigate when mounted through the cloud. In some cases, the amount of time it takes to enumerate the contents of a modestly-sized folder is far in excess of the amout of time it would take to ssh in, cd to the folder in question, and type in "ls -lha".

Worse, as these mount processes tie into the OS' lower-level I/O stuff, it can cause complete freeze-ups of Explorer/Finder and in a few cases necessitate a complete reboot.

Can anyone explain why this is so slow? I am interested more in knowing why, and less in a practical fix, though if there is anything I can relay to the server guys I'll gladly send it over.

Thanks!

Tom

Best Answer

Using SFTP (or similar) as a distributed filesystem is using it for something it's not designed for. The ability to cache, read ahead or detect network disruptions is dependent on the quality of the implementation of the client, which has to do with an underlying protocol not designed for this purpose.

If you want better performance, using a proper distributed filesystem such as NFS or maybe AFS (this one is designed for local caching but generally harder to install) should help.

Regarding the client's performance, you might want to try something like sshfs (via MacFuse) on the Mac for example to see if it works better for you.