I have two machines - a 1.2G and a 1.3, which are running qwsv, Quake
server processes.

qwsv: BSD/OS i386 compact demand paged executable

from the 1.2G machine ... which is running two of them...

quake% ps -alx | grep qwsv
  666  3759  3586  12   2  4  7876 11220 select SN   ??   21:25.51 ./qwsv -port 28000 +setmaster 192.246.40.12:27006
  666  5481   144  12   2  4 24624 12252 select SN   ??  129:05.72 ./qwsv +setmaster 192.246.40.12:27006
quake% ps -alx | grep qwsv
  666  3759  3586  12   2  4  7876 11220 select SN   ??   21:25.52 ./qwsv -port 28000 +setmaster 192.246.40.12:27006
  666  5481   144  13   2  4 24624 12252 select SN   ??  129:05.73 ./qwsv +setmaster 192.246.40.12:27006

Only shows a change in the last digit of the CPU time, (probably no one
using the servers right now)

quake% vmstat
 procs   memory     page                    disk faults   cpu
 r b w   avm   fre  flt  re  pi  po  fr  sr w0   in   sy  cs us sy id
 1 0 0253864  1932    2   0   0   0   0   0  0  253  839 177  7  2 91

And lots of idle CPU.

on the 1.3 machine...

quake2% ps -alx | grep qwsv
  666   897   895   0   2  0 24588 11040 select S    p0-   7:06.57 ./qwsv 
quake2% ps -alx | grep qwsv
  666   897   895   0   2  0 24588 11040 select S    p0-   7:06.58 ./qwsv 

Likewise, CPU isn't racing upwards....

quake2% vmstat
 procs   memory     page                    disks      faults   cpu
 r b w   avm   fre  flt  re  pi  po  fr  sr w0 f0 m0   in   sy  cs us sy id
 0 0 0 87548  8308    3   0   0   0   0   0  0  0  0  110 46754  10  9 29 62

but 62% idle time, again no one using the server right now, but during 
use it dropped to 2%, while the 1.2G machine stays above 85% with _two_
qwsvs running. Top also disagrees - says 99%+ idle on both machines, but
on the 1.3 server people were complaining of noticeable pauses in
gameplay, and I saw that the CPU time for qwsv was counting fast.

Both machines run not much else but these (the 1.2 has a third, non-quakeworld
quake server, but it doesn't get much usage.), top consistently reports
the qwsv processes as the cpu pigs.

Any changes in select between 1.2G and 1.3? Or perhaps in BSD emulation?
Can anyone suggest why vmstat and top disagree so widely? Vmstat seems to
be the 'right' measurement based on the complaints.

						David Maxwell
						david@vex.net