Re:Re: R14.1 vs. R16 ADS performance

Discussion created by ca.portal.admin on Mar 4, 2006
Hi Chris -

We are OS/390 V2R10.9 running on a 9672 - R24 Processor.
I know this sounds a little ancient, but our mission is to get off the
We are slow in getting there, so there is some renewed interest now in
upgrading and being on on supported software.

We are trying to move from R14 - 9810 (plus many, many APARS) to R16 -

The COBOL is COBOL for MVS and VM - Release1.2
COBOL isn't a factor in the sample program we've been using.

The sample program is Assembler, because I didn't want to have answer a
of grueling questions about LE differences.
In fact, the Assembler program is a modified version of the sample
in the Assembler DML reference manual.
One of the modifications was a ""Loop Limit"" to say how many times we
to area sweep the ORG-DEMO area.
The latest round of tests have this Loop limit set to 50,000.

The DEMO DB is the standard tiny DB, and the buffer I have is 495 pages.
So there is only I/O to load the areas once and then everything after
are ""finds in the buffer"".
The BUFFERSTAT report shows this is the case.
That's another area I didn't want to have to answer a lot of questions
# of physical I/O's, type of DASD, tuning of I/O subsystem, etc.

To remove another variable, I used DD overrides so that both R14 and R16
tests would go against the same DEMO database.

Anyone who wants a copy of the sample program I am using is welcome to
Just drop me a line.
Those of you making the leap to R16 might want to run it to see what
differences you are getting.
I feel alone here because, intitally, C.A. had said they have gotten no
other complaints and there was nothing significant that would cause such
increase in CPU.

Both C.A. and I are testing with this same program, although the CPU
increases they are seeing aren't as dramatic as mine.
But, then again, they have been comparing R15 to R16 for most of their

I think those of you that saw little or no increase are running programs
that do little database work relative to other CPU-crunching business
You probably have big-iron monster machines that are masking most
differences, too.

The sample program is all DB work.
So the differences really stand out.

Jon Gocher

----- Original Message -----
From: ""Chris Wood"" <Chris.Wood@GOV.AB.CA>
Sent: Friday, March 03, 2006 11:01 AM
Subject: Re: R14.1 vs. R16 ADS performance


Can you tell the rest of us what OS and hardware, version of COBOL and
other tuning methods you may use for the test job and other jobs please?

Chris Wood
Alberta Department of Energy

I sent this message earlier but have not received the reply from the
Listserv yet so I don't believe that it was posted.