Re: RPG IV Performance -- RPG400-L

Lisa, et al,

In a message dated 1/19/01 8:15:37 AM Eastern Standard Time,
Lisa.Abney@universalflavors.com writes:

Hi all! I've just heard some rather negative performance things on RPG IV,
and
wonder if anyone can give me some feedback on how true this might be.

We're working with a consulting company who is doing some performance
analysis
on some of our programs. They seem very knowledgeable, and I have a lot of
confidence in what they've done up until now. However, today they were
showing
us a mock-up of a trigger program they want to use. As they explained it,
this
trigger program will be constantly running in the background for every user
on
the system to monitor changes to two files, and will feed data to a dataque.
The program they showed me was written in RPGIII, and I made my usual
request to an outside contractor that this be done in RPGIV. His response
was "Sure, if
you want the program size to be 5 - 10 times the size of an RPGIII program."
When I asked him to explain that, he only said that, in his experience,
this is
always true, and that it would have a very negative performance. I even
mentioned removing observability (not that I really understand what that
means,
but I just read something the other day about that being a way to reduce
program
size!), and he said that might move it down to 3 - 5 times the size of an
equivalent RPGIII program. The program will only be about 50 - 100 lines of
code.

Can someone explain if this is true, and, if so, why? And, if true, what
does
this really mean from a performance standpoint?

To reiterate some points already made, and to highlight some that weren't.
Program size has absolutely no impact on it's performance. I could write an
absolutely enormous program in RPG III that, given a single unexecutable IF
statement skipping 98% of the logic, would still perform quite well.
However, comparing apples to apples, RPG IV will almost always be larger
given the same functionality. This is a case where size doesn't matter.

Now to the unmentioned! Unless you're running SETOBJACC against it, a
trigger program is _NOT_ constantly running in the background. It is invoked
as part of whatever job "triggers" it by performing the proscribed change
action against the relative file. Unless the only thing that every single
one of your users is doing is constantly updating one or both of these two
files at the same time, the trigger is not running at all times. The trigger
also does _NOT_ write to a data queue unless that's how you wrote it.

It sounds to _ME_, with the information given, that one of two things is
happening here:

1. Simon and Mr. Peabody have stepped into the "Wayback" machine and we're
rehashing ILE performance on CISC in the days when there weren't many RISC
machines on the market (see the archives from three or so years ago). There
was no performance advantage indeed, a degradation, of ILE on CISC.

2. These "consultants" do not understand true UDB/400 triggers, and have
written some sort of "NEP" solution involving read waits and data queues
instead of utilizing the native "TRG" commands.

The more I think about it, the more the latter seems to be the case. I
suppose that the "acid test" would be to ask your consultants which AS/400
command adds a trigger to a file, and which removes it. During add, what is
the significance of the ALWRPTCHG parameter?

Finally, I don't know what we're talking about in terms of "lines of code",
but I have yet to write a trigger that required more than 20 "C" specs. If
you need more than that, you either need to normalize your database or
reconsider why you're using triggers...

JMHO,

Dean Asmussen
Enterprise Systems Consulting, Inc.
Fuquay-Varina, NC USA
E-mail: DAsmussen@aol.com

"Without a struggle, there can be no progress." -- Frederick Douglass