Could you tell me what does it mean load average in Uptime command ?
Where could I get uptime algorithm ? I want know how this values are extracted
from the system.

Why vmstat column 1 show me one value and uptime load average show me other ?

Running queue and load average are the same metrics ?

uptime
  7:21pm  up 30 day(s),  2:15,  7 users,  load average: 8.46, 8.17, 9.50 

vmstat 1 10
 procs     memory            page            disk          faults      cpu
 r b w   swap  free  re  mf pi po fr de sr s0 s6 s3 sd   in   sy   cs us sy id
 2 2 0 37306896 20746832 495 6715 4198 14 10 0 0 10 0 9 12 7036 2542 2107 205 48
 0 3 0 39240048 22500248 1918 1225 8600 0 0 0 0 0 0 0 2 1403 21316 5048 30 6 64
 0 3 0 39235224 22497800 1835 826 8864 0 0 0 0 0 0 0 2 1387 17596 4534 29 2 69
 0 4 0 39239216 22498872 1637 343 13720 0 0 0 0 0 0 0 50 1361 15085 4225 34 2 64
 0 4 0 39236136 22495568 588 96 8096 0 0 0 0 0 0  0 20 1431 15901 4588 31 2 68
 0 5 0 39235720 22495408 1456 1647 7752 24 8 0 0 3 0 3 0 1752 21734 5503 31 4 66
 0 5 0 39243648 22502472 7811 3190 9800 0 0 0 0 0 0 0 70 1917 44539 9057 43 10 7
 0 5 0 39238280 22498760 9020 43 12520 0 0 0 0 0 0 0 11 1652 35099 7837 41 6 53
 0 5 0 39237984 22498208 6412 844 11472 0 0 0 0 0 0 0 4 1839 19369 6058 37 4 59
 0 2 0 39265408 22523344 1253 1125 9032 0 0 0 0 2 0 3 1 1095 17738 5229 38 3 59

On Tue, 4 Jan 2005, Shiva MahaDeva wrote:

> Could you tell me what does it mean load average in Uptime command ?

Very briefly, the load is the number of running and runnable
processes, and the load average is this number averaged over
a given time span.

I explain this in more detail in my book, Solaris Systems Prorgamming.
You can find the details, including the sample chapter that conveniently
has the bit on load avereages, on the book's site: www.rite-group.com/rich/ssp .

> Why vmstat column 1 show me one value and uptime load average show me other ?
>
> Running queue and load average are the same metrics ?

Not quite.  The load average is what I described above, but see the
vmstat man page for what its columns mean.

HTH,

-- 
Rich Teer, SCNA, SCSA, author of "Solaris Systems Programming"

                                                    .  *   * . * .* .
                                                     .   *   .   .*
President,                                          * .  . /\ ( .  . *
Rite Online Inc.                                     . .  / .\   . * .
                                                    .*.  / *  \  . .
                                                      . /*   o \     .
Voice: +1 (250) 979-1638                            *   '''||'''   .
URL: http://www.rite-online.net                     ******************

Note that load average prior to Solaris 10 was calculated using the
number of threads on the
run queues (plus any on cpus) sampled 100 times a second at the clock
interrupt.  As most
engineering students know, sampling a continuous function at discrete
intervals only yields
accurate results if the sampling frequency is 2x the highest signal
frquency.  Since 20000 threads can come and go in a single clock tick
(esp. on large jvms ), it's easy to find
machines with a  load average of 100 and 85% idle cpu :-).  Sort of
counter-intuitive.

We now aggregate time spent on the run queue for each thread as part of
the microstate
accounting cleanup;  we now correctly account for time both on cpu and
on the run queues,
irrespective of when the clock thread runs.  No more hiding behind the
clock tick!

- Bart