Re: Dynamically allocated tables and %DEALLOC

Another truism that makes the rounds is that a Load All subfile takes

longer than a page-at-a-time subfile.

Good point, though I have taken the page at a time approach simply because
data sets are being used so much more outside of the box that it warrants
thinking about how the data is being used.

The other thing to consider in all of this is how the amount of users can
affect performance. Sure, when we are developing an application it seems
fast, but once a couple hundred users are actively loading that same subfile
it could bring a system to it's knees. I had this happen to me with a
JavaServerFaces page where I thought "ah, they will never have more than
1000 records in that table...". Needless to say they far surpassed that
number with a new business requirement and I had to re-work the JSF
"subfile". Once I had the infrastructure built for doing page-at-a-time
logic it was easy to implement in other new or modified programs. So now my
practice is to never to the "load all" approach, because the expense to go
back and fix it (ie. relearn how the apps works and make sure I am not
stepping on toes with the change) far outweights the cost of doing it right
out of the gates.

It may seem like I am wandering from my original assesment, but in the end I
am just trying to bring up areas that sometimes make sense and other times
not. YMMV.

Aaron Bartell
http://mowyourlawn.com

On Mon, Aug 18, 2008 at 7:59 AM, Jerry Adams <Jerry@xxxxxxxxxxxxxxx> wrote:

Aaron,

I wonder the same thing about this (activation groups) and some other
similar things that we learned over the years.

I've never noticed (V5R4, Model 520) a new, named activation group
impacting performance.

Another truism that makes the rounds is that a Load All subfile takes
longer than a page-at-a-time subfile. Undoubtedly this is true, but is it
noticeable? I've discarded page-at-a-time and use strictly load all
(haven't had to worry - yet - about the 9,999 record limit). I've loaded
+5000 records in sub-second response time. On our old iSeries (720) I never
tried that. On a B10 I'm pretty sure it would have been noticeable, but
always did page-at-a-time until a few years ago so I have no comparisons. I
just know that on the newer hardware, it is perceptually not noticeable.

Jerry C. Adams
IBM System i Programmer/Analyst
B&W Wholesale
office: 615-995-7024
email: jerry@xxxxxxxxxxxxxxx