[plug] file transfer performance linux/windows and filezilla

[Denis Brown]

On Tue, 11 May 2004, Craig Ringer wrote:

> Denis Brown wrote:
>
> > Hmmm... no evidence of a config file in my .ssh directory.   Will dig a
> > bit deeper later.   Thanks for the tip.
>
> It's not created by default, but it's a useful thing to keep. Mine is
> attached, in case it's of any interest. 'man ssh_config' for info.
>
Well, I've had a few mintes to look at it this evening.   FWIW here are
the results so far...

a)   Set up /etc/ssh/ssh_config global config file to use blowfish as the
standard algorithm on the two Linux machines previously used for testing.
b)   Transferred an .iso file 650MBytes, same one I used earlier.
c)   Transfer time with default cypher (3des???)  59 seconds
     Transfer time with blowfish cypher  58 seconds
     Transfer time with -c blowfish command line option, 58 seconds
and interestingly...
d)   Transfer time from machine to itself, 1 minute 12 seconds!

For the latter test I used
    scp user at machine:file.iso user at machine:file1.iso
thus aiming to elimnate the network aspect.

While the transfers were in progress I ssh'd in and ran "top" to examine
the status quo.   In the case of transfer from slower machine to faster
machine, ssh was running around 81% of cpu.   In the case of transfer
slower machine to itself, ssh was at 44.7% and sshd was at 32.5% which
would explain the sloth (poor machine was doing both ends of the
transfer!)

Correct me if I'm wrong but the above would seem to vindicate the network.
So begs the question why ssh (or Debian's implementaion of ssh???) is so
slow.   I think it was Shayne who commented that ssh should outperform ftp
due to ftp's overheads.   Bill earlier mentioned the cpu load and that
would certainly seem to apply here.   Unless there are howls of derision
<smile> I will have a fresh look tomorrow, at why my ssh is slow.   I'll
Google at the least.

Just to recap...  several Windows NT4 machines (running FileZilla) are
equally lethargic for transfers when connected to several Linux (Debian)
machines, two of which should not lack for horsepower.   So it would seem
not to be a simple case of one client, or server, being ill or
ill-configured.

Not yet tried:
i) transfer from host to same host but using localhost ip 127.0.0.1
ii) transfer from one host to another using ip numbers rather than dns
names in the scp command.
Based on tonight's experiments I don't think either of these
straw-clutchings would make much difference.

Cheers,
Denis