[log in to unmask] wrote:
>
> Re:
>
> > they start impeding on some resource.  We have eliminated the Put/Delete
> > semaphore by having each batch job write to a different database.  Disk and
> > memory utilization is very low so I do not believe that we have a bottle
> > neck there.  Global CPU is only at 40% during this time.
>
> What does a performance tool like SHOT, SOS, PROBE, GLANCE, etc., show
> the processes waiting on?
>
> --
> Stan Sieler                                          [log in to unmask]
> P.s.: please forgive typos/brevity,      http://www.allegro.com/sieler/
>       I'm typing left-handed for awhile.

Hello Stan,

Well brevity is a virtue but I hope you are better soon.

This Image thing is a puzzler:

1.  No masters.
2,  TPI off.
3.  Old data base purged and rebuilt, I assume.
4.  I assume also an efficient blocking factor.
5.  No sorted chains, I would also assume.
6.  Is access to the root file required for each put?  I think not
    but even if it were it would stay in memory, so still no problem.

Is there a kind of bug in Image where the code uses the same semaphore
even
though different data bases are involved?  I am grasping at straws here,
I know.  I wonder what the test performance is like when only one
data base is rebuilt at a time.

Nick D.