Hello,

and thanks for all the replies I received.

After weekend, I'm "back by the forge", so it's time to continue
on network performance.

To summarize the many questions:

1) "Can you easily test driver patches on your SS20 to see if they
improve performance for you?"
	-Yes, I can do that: the machine serves as a production
	firewall, though, but there's not much traffic going
	through at this time of year (summer holiday), so
	it's not a problem.

2) Will try a new kernel with nsphy enabled

3) The CPU usage during test (534MB file from one side to
another), and other statistics, taken from "systat -w 5 vm":

(somewhat edited for readability)

...
          memory totals (in KB)
         real   virtual    free
Active   5748      5748   7089
All     22588     22588  406812

Proc:r  d  s  w       Csw  Trp  Sys  Int  Sof  Flt
           3           15    1   34 3250         1

   0.2% Sy   0.2% Us   0.0% Ni  81.8% In  17.8% Id 
|    |    |    |    |    |    |    |    |    |    |
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%          



 Interrupts
3239 total
 208 lev1
     lev4
     lev6
2831 lev7
 100 clock
     lev12
 100 prof

...

The reason I started wondering, I had two Big Mac's
(be) on it before, and the performance was *exactly*
the same. Now, switching to two hme's, there's no
difference in performance whatsoever..

Also, if someone would have time to take a look on
my ipmon problem I posted to port-sparc64, I could
switch this one over to Ultra-1 ;) I think it's ipmon
failing silently on file descriptor, diagnosed from
ktrace output..

Regards,

Tomi

PS: Please reply to the mailing list, as my e-mail address
is obfuscated to repel spammers.