Le 18/11/10 04:46, Peter Stuge a écrit :
>> Presumably the difference there is due to packet delays.
> Personally I don't really *know* what the problem is, so to profile
> this will require looking a lot at timing and packets, both from
> inside the libssh2 app, and from another viewpoint outside the app,
> e.g. sniffing packets directly from the OS, or even on the wire.
I'm 100% sure it's due to packet delays. Mark, as a suggestion, could
you provide us a little more information, like the RTT between the two
hosts (even on the LAN), and the sample of code you used for the
benchmark (so it's reproducible). We all know that performance of a lib
may be very well depending on the way it's being used.
I'm also interested in having libssh's results on your site as well
(both 0.4.6 and git master).
> For these upload tests with libssh2, where did the data come from
> that you uploaded? If there is any other I/O in the _sftp_write()
> loop then please pre-load all data and remove any swap file, so that
> all data is in fact available directly. Again the purpose is not to
> significantly improve the numbers, but to create an easily repeatable
> environment for further testing and development.
I second this advice. "cat file > /dev/null" before running the test may
be enough. On remote side, when possible write to /dev/null or to a SSD
Received on 2010-11-18