James Wilkinson wrote:
James Pifer wrote:
I used scp and the results were rather disappointing. Using scp I had
the following results:
11Mb/s - scp would go around 815K
54Mb/s - scp would go around 1.1MB
10/100 NIC - scp would go around 6.4MB
General note: scp is great for real-life use over unprotected networks.
But because there is the overhead of encryption and decryption, scp can
be limited by processor power at one end or the other of a link, rather
than raw throughput (especially if the network card itself requires a
lot of processor time to do its stuff). This means it's not the best
choice for measuring anything other than scp throughput.
James.
It has been my experience that the overhead of encryption for SCP does
not come into play in the standard distribution of OpenSSH, but rather
is limited by the SSH buffer sizes. Using the HPN-SSH patch
(www.psc.edu/networking/projects/hpn-ssh/), I've gotten exponential
increases in scp throughput...
--
Ed Kim, RHCE
http://www.rhatbox.com
