Date:         Thu, 3 Nov 2005 13:44:59 -0500
Reply-To:     Charles Harbour <charles.harbour@PEARSON.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Charles Harbour <charles.harbour@PEARSON.COM>
Subject:      Re: z/OS memory questions
Comments: To: Alistair.Gordon@NEWTYNE.COM

Yes, task memory is what was actually used, total memory is what your
allocation is.  In looking back at your original posting, you would use the
total memory allocations to tune your memleave option, to possibly take
advantage of unused program memory, and allow it to be used for data.  But
your test was so small, it never approached the allocation limits.  And
that part really is critical--until you start seeing swaps out to disk, you
really can't measure what the memory tuning accomplishes.

And, as I posted earlier, it really depends on the data and the program
itself.  If you are <really> serious about tuning (assuming an extremely
large dataset and stable program), you could look at using a pre-compiled
program.

In discussions with the folks at SAS, Inc., there are some modules that are
not optional--it's going to load ODS, even if you never call ODS.  So, from
an 'extreme programming' sense, you are limited to how the language
operates.  But you have plenty of tuning knobs at your disposal with the
bufsize/bufno and memsize/memleave options.  And don't forget to crank your
region up to the system max.

CH


On Thu, 3 Nov 2005 09:37:13 -0800, Ali G <Alistair.Gordon@NEWTYNE.COM>
wrote:

>Thanks for replying Charles.
>
>Another question then:
>
>With Fullstimer I am looking at TASK memory and TOTAL memory.  Is TASK
>memory the better identifier of memory usage for a step i.e. that's how
>much memory was used in the private area ?
>
>Regards,