SSH: Single Threaded problem in T2000

Question

We are running into a problem where the standard OpenSSH that comes with Solaris 10 is extremely slow on the T2000s (on the primary and LDOM both). Here is an example of a timed rcp and scp of a 139mb file:
# time rcp ocp-engine.log psldom01:/tmp
real 0m4.10s
user 0m0.00s
sys 0m1.53s
# time scp ocp-engine.log psldom01:/tmp
ocp-engine.log 100% |*****************************| 139 MB 00:35
real 0m37.16s
user 0m5.53s
sys 0m1.85s
The official word we received from SUN on a case we opened was "Yep....use FTP and Rservices". We are playing with the following for a solution: [http://www.psc.edu/networking/projects/hpn-ssh|http://www.psc.edu/networking/projects/hpn-ssh]. However, the future of Sun in our environment does not look good if we can not resolve this. VMWARE has already replaced the SunRays.
Has anyone run into this problem and what have you done to deal with it?
Frustrated,
James Council

Answer 1

I don't have any of the R services enabled, but I was able to do a similar test using FTP in comparison to SCP:
With FTP:
ftp> put Net*
200 PORT command successful.
150 Opening BINARY mode data connection for NetBackup_6.5_CLIENTS3 - Solaris BSD and IRIX.tar.
226 Transfer complete.
local: NetBackup_6.5_CLIENTS3 - Solaris BSD and IRIX.tar remote: NetBackup_6.5_CLIENTS3 - Solaris BSD and IRIX.tar
375810048 bytes sent in 47 seconds (7811.07 Kbytes/s)
With SCP:
[root@t5240:scripts, Tue Jul 08]>time scp NetBackup_6.5_CLIENTS3\ -\ Solaris\ BSD\ and\ IRIX.tar [email protected]:/tmp
Password:
NetBackup_6.5_CLIENT 100% |**********************************************************************************************************| 358 MB 01:19
real 1m23.789s
user 0m35.107s
sys 0m8.469s
SCP with compression was even worse (I know you weren't using it):
real 3m21.334s
user 3m3.615s
sys 0m8.575s