On Mon, May 24, 1999 at 02:26:41PM -0400, you [Patrick Farley] claimed:
> Did some further testing.
> These are the kilobytes per second that I was getting when transfering between
> different machines on the same subnet.
>
>
> Version SSHD -> 2.0.13 1.2.27 1.2.26
> SCP
> 2.0.13 53kb/sec 38kb/sec 500kb/sec
> 1.2.27 40kb/sec 36kb/sec 40kb/sec
> 1.2.26 550kb/sec 40kb/sec 580kb/sec
>
> I transfered the same 4 mb file between all the machines. Each machine has a
>similar sshd_config and ssh_config.
> The times aren't exact, but there is such a difference between them accuracy should
>matter that much.
Basicly, I think you have two ways to go: diff the source of 1.2.26 and
1.2.27 (the socket/scp related changes should not be that numerous)
and/or use tcpdump.
If I were you, I'd try the first option first. If you find a
suspicous-looking change from 1.2.26 to 1.2.27, just undo it in the
1.2.27 tree and try if that remedies the problem.
On pc-platforms that I use ssh on, I've had no performance problems
(other than that piping Linux smbfs through ssh is damn slow... ;) ), and
I haven't tried 2.0.x either, but I hear these kind of scp-performance
problems are common with 2.0.x. So it could be beneficial to spot the
problem.
-- v --
[EMAIL PROTECTED]
