RE: V5R1 slow? -- MIDRANGE-L

> Pat Barber:
>Do you have Performance Tools(and know how to use it ??)

Not afraid to ask the tough questions, are you Pat?  :)
We have Performance Tools.  I'm about as dangerous with them as I am with my
circular saw.


>It is never odd for performance hits on upgrading to a new
>release in my opinion. You need at least a week or two of full
>production for the machine ro "settle down" and "uncompress" about
>a zillion objects that came compressed when you loaded them on.

>Wild guess(s) is not the way to handle a box of that size and
>based on what you saying, you upgraded for a reason ??? Did you
>add any hardware when you made this upgrade(like more memory) ??

We bought the biggest baddest AS/400 money can buy so that we *could*
function exclusively on wild guesses.
We saw performance hits in previous upgrades, but nothing like this week's
performance crisis.  We upgraded to V5R1 only to stay current.  Last year we
put off V4R5 'til the last minute and regretted it.  Upcoming projects may
make it impossible to upgrade later in the year.

We added hardware during the upgrade (cannibalized from an even more
obsolete 650 12-way) but it was just disk (more on that later).  Buying any
parts for these machines is a serious waste of money.  The moment you
install anything it's worth as little as the rest of the machine.


>I guess things have changed when a million dollar box is considered a
>dinosaur....

I watched as the value of our other million dollar box (the 650 12-way) sunk
into the low five digit range.  Upgrade paths are a wonderful thing.

I appreciate everyone's input and help.  There's been a chorus of folks
jumping to conclusions and wanting to blame the new release.  It was getting
in the way of productive analysis.  Here's what really was wrong:

We added a couple new (old) drives from the ol' 650 at the same time as the
upgrade, and never bothered to balance them into the ASP ("there wasn't
enough time").  In the past it hasn't been a big deal because we usually
were adding hundreds of gig.  This time we added two RAID sets of 8 disk.
Immediately after the upgrade and disk one of the first batch processes
reorganized a few huge files including a very active 34 GB file and its
logicals.  The reorg wrote to the new empty disk.  So, the rest of our 800
GB is spread over 90 or so drives while all access to one table is spread
across 16 disks in the same two RAID sets.  All the catch-up work from the
weekend is trying to read or update the same critical files in a contentious
configuration.

I've tried to ASP balance on usage.  It didn't do what we needed.  Now I'm
up all night performing a save and restore via save file, with ASP balance
on capacity in-between.

You've all helped me to confirm that there are no smoking guns in V5R1
(other than the ones Al Barsa will tell you about), and allowed me to focus
back on the obvious.  Tomorrow I'll be looking into some of Philipp Rusch's
and Adam Driver's points.

Much, much thanks...

-Jim

James P. Damato
Manager - Technical Administration
Dollar General Corporation
<mailto:jdamato@dollargeneral.com>