Re: Auto Tuning

[Simon Coulter <shc@xxxxxxxxxxxxxxxxx>]

On Wednesday, February 4, 2004, at 09:16  AM, Vern Hamberg wrote:

> The odd thing to me is that no job runs longer than the internal time 
> slice, but the activity level is held for it until the end of its 
> external time slice value, or a long wait.

Perhaps I'm misinterpreting what you wrote above but the previously 
quoted IBM document (which was badly worded and seemed confused by its 
own numbers) said that if the job has a longer time slice than 500ms 
and reached the 500ms boundary (indicating it has not completed its 
work) then the CPU would be given to another job but the activity level 
of the first job would be held (and its resources remain in main 
storage) because the system knows that job has more work to do and it 
is still within the user-specified time slice so the job shouldn't lose 
its activity level. The activity level is not held if the job has 
nothing to do even if it has a long time slice (although job resources 
may stay in main storage if main storage is not overcommitted).

It should have been worded as: The activity level of a job is held 
until the job reaches the external time slice or a long wait occurs 
even though an internal time slice occurs every 500ms (or whatever 
actual value is used by current hardware) and the processor may be 
given to another job.

Regards,
Simon Coulter.
--------------------------------------------------------------------
   FlyByNight Software         AS/400 Technical Specialists

   http://www.flybynight.com.au/
   Phone: +61 3 9419 0175   Mobile: +61 0411 091 400        /"\
   Fax:   +61 3 9419 0175                                   \ /
                                                             X
                 ASCII Ribbon campaign against HTML E-Mail  / \
--------------------------------------------------------------------