Hi!

Does NetBSD ever swap or is that a forgotten art?

I've been trying to run more things than fit in memory and
I think NetBSD behave in a very suboptimal way.  Even when
the free memory goes to 0 it refuses to completely swap
out a process and let the rest of them make progress.

All code that concerns swapping seems to be surrounded by 
'#ifdef OMIT' and similar things.  This leads me to assume
that swapping is now a lost art.  Too bad, performance degrades
incredibly when there isn't enough memory.  (I know, I know,
you can buy more memory :-)

Setup: two process each wanting around 10M of memory
running on a 60MHz Pentium with 16M memory, using
NetBSD 1.0_BETA.

Below is an excerpt from 'vmstat 1' and 'vmstat -s'.

*BSD Unix used to be (in the good ole vaxen days) good at handling
more processes than it had memory for.

(FYI, Solaris 2.3 is about as bad as NetBSD at this test
whereas SunOS 4.1.3 does much better (but noot good),
and Dynix does really well.)

	-- Lennart

vmstat 1:  Notice, the free memory is 0 for a long while.

 procs   memory     page                    disks         faults      cpu
 r b w   avm   fre  flt  re  pi  po  fr  sr ?0 ?1 ?2 ?3   in   sy  cs us sy id
 0 9 0 66724   204    2  12   1  61   0  99  0  0  0  0    1   26   6  0  0 100
 0 9 0 66724   224    7   1   6  61   0  71  0  0  0  0    1   26  10  0  0 100
 0 9 0 66724   336    6   4   5  44   0  79  0  0  0  0    1   26  54  0  0 100
 1 3 0 66724   172  430 177 160  76   0 839  0  0  0  0    1   40 114 26 23 51
 0 4 0 66724     0   46   1  44  16   0  16  0  0  0  0    1   28  10  5  0 95
 0 4 0 66724     0   14   1  12  58   0  69  0  0  0  0    1   28   9  1  0 99
 0 4 0 66724     0    5   2   4  56   0  60  0  0  0  0    1   28   9  1  0 99
 0 4 0 71500     0    3   1   2  59   0  60  0  0  0  0    1   52  13  0  0 100
 1 4 0 80768     0    7   4   6  57   0  65  0  0  0  0    1   32  10  1  0 99
 0 6 0 85348     0    4   1   1  60   0  60  0  0  0  0    1   31  13  0  0 100
 0 6 0 85348     0    3   1   2  57   0  58  0  0  0  0    1   28  11  0  0 100
 0 8 0 85348     0    6  31   9  58   0  98  0  0  0  0    1   28  11  0  1 99
 0 7 0 85348     0    6   9   3 339   0 348  0  0  0  0    1   61  38  0  0 100
 1 4 0 85348   272   25   7  24  48   0 148  0  0  0  0    1   52  95  1  1 98


vmstat -s: Notice, no swaps.
(I had a quick look in the source and the vm_swpout counter
does not seem to be incremented anywhere.)

  1422835 cpu context switches
        0 device interrupts
   917468 software interrupts
  4033376 traps
  6138877 system calls
  3172220 total faults taken
        0 swap ins
        0 swap outs
        0 pages swapped in
        0 pages swapped out
   424793 page ins
   241135 page outs
   348521 pages paged in
   237964 pages paged out
   831677 pages reactivated
	...