RE: We've Added more memory...but I can't remember!

Charly,

Thank you for the correction.  I was being overly simplistic to the
point of inaccuracy regarding the machine pool.

On faulting levels, I'm going to give that some thought.  Somewhere in
the continuum between no faulting at all and thrashing, there is a point
at which the addition of more memory is no longer cost effective.
Perhaps it is a moot point because they no longer publish guidelines.  I
tried to find current information on the InfoCenter, but it seems that
IBM's tactic here is to direct you to the measurement tools.  Somewhere,
embedded in the code and tables for those tools are some guidelines, but
they are not putting them out in readable form (that I could find).

Thanks,
Andy

> On Behalf Of Charly Jones
> Subject: RE: We've Added more memory...but I can't remember!
>
> > > If it does, any GENERAL rule of thumb to follow for incrementing
the
> > > MACHINE pool?
>
> Rick  --
>
> I disagree with Andy on a couple of points.  First, unless IBM has
made
> some
> major architectural changes that I don't know about, the VAT (virtual
> address translator) mechanism requires some pinned memory in the
machine
> pool to keep track of what "real" address is stored in each memory
frame.
> If I remember correctly - when you add 16 gigabytes of memory to a
system
> you need to put 1 gigabyte of additional memory in the machine pool
just
> for
> that purpose alone
> Second, the "guidelines" for acceptable faulting are totally bogus.
It is
> not "OK" to have hundreds of non-database faults per second.  Every
fault
> requires something to be brought from the disks into memory.  These
poor
> disks are busy enough (don't get me started on the disk problem.)  If
you
> have a performance problem you can either keep buying hardware until
the
> performance improves enough, or you can take action to reduce the
> unnecessary faulting.
>
> --  Charly
>
>
> > > On Behalf Of Rick Rayburn
> > > Subject: We've Added more memory...but I can't remember!
> > >
> > > ...if I need to "goose" up the machine pool with additional
"wattage".
> > >
> > > the memory was added because we got a great deal on the chips NOT
> >because
> > > we
> > > were experiencing problems. I believe all of the additional "K"
was
> >dumped
> > > into *BASE but I'm not certain.
> > > Does anyone remember/know if memory additions ALWAYS dump into
Base?
> > > If it does, any GENERAL rule of thumb to follow for incrementing
the
> > > MACHINE
> > > pool? I believe I OVER-allocated memory to the "SPOOL POOL" by
> >granting an
> > > average of 300 K per active writer. Any thoughts on that as
well...or
> > > anything else memory-pool related?
> > >
> > > Thanks all.
> > >
> > > Rick Rayburn
>
> Charly Jones