Re: High NON-DB faults in machine pool (~10) on V5R2 w/1Gig in pool

["Steve Landess" <steve_landess@xxxxxxxxxxx>]

From: "Walden H. Leverich III"
> We just moved several machines to V5R2 (720 & 730) and we're experiencing
> high (10-15) non-db faults in machine pool on the 730. Auto-tune is on and
> the machine pool at 1 Gig (1064 meg to be exact) out of the total 6 Gig
> memory -- there is 1.5TB of disk.  Can this be right? We'd consider
turning
> off autotune, but should we have more than a Gig in the machine pool?
>

Walden -

I have worked with several of the AS/400 autotune facilities and observed
what they do to keep the system tuned, as well as the IBM perfornamce
adjustment (QPFRADJ).

I know that there are a _lot_ of variables involved, but in my experience I
have usually seen 15 % to 20 % (sometimes even more) of the system memory
needed in the machine pool.   So, given this example, I think you may have
enough memory there (1/6 = 16.7 %), but there is a formula in the Work
Management Guide for calculating the amount of memory needed.  Here is a
link to a page in the Infocenter where you can view the PDF version of the
V4R4 Work Management Guide (the title on the web page says V4R5, but is it
wrong):

http://publib.boulder.ibm.com/iseries/v5r1/ic2924/info/rzaks/rzaks1.htm#manu
als

On page 242, there is a formula for manually calculating the size of the
machine pool.  If you haven't read this book, I would read it cover to
cover, particularly chapters 4 (Subsystems) and 14 (Performance Tuning).

HTH

Steve Landess
Austin, Texas
(512) 423-0935