RE: Alternate Platform for RPG

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> From: Bartell, Aaron L. (TC)
>
> There was no change in the amount of time it
> took to process with the Servlet compared to the RPG program.
> Does that say something?...  IMO, yes.

It depends upon the data source.  You were reading data from somewhere,
right?  The fact that you were showing no CPU footprint means that you were
disk bound, and a disk bound program is invalid as a measure of CPU
performance.  Benchmarks are a very subtle thing, and eyeball measurements
often don't tell you what they seem to at first glance.


> If nobody notices that a process is running
> slower, and it doesn't leave a big CPU footprint, then I would
> consider that
> successful.

If a program runs in a forest and nobody complains, is it performing well?
Not necessarily.  If they're data source constrained, one could be
performing ten times as well as the other, but since they're both spending
the majority of their time waiting, there's no comparison.  The closest you
can come in that case is to compare actual CPU time used, which is available
in WRKACTJOB.


> Do you have any numbers for Java vs. RPG where I/O wasn't
> involved?  I would
> like to see those because my main use for Java doesn't involve
> any Database
> I/O.

I did quite a few, but as I said, I think they need to be revisited.  I'll
work on that in my CFT (Copious Free Time).


> Or at least I haven't found a use/need to do that yet.  Also, when
> does the time come where you utilize new tools to make development simpler
> at the expense of a couple CPU cycles?

Different question!  And I have to KNOW how much performance I am giving up
before I can make that business decision, right?  So in order to make a
truly informed decision, it's CRUCIAL to know exactly how well one technique
performs against another.


> Joe, I don't think you would argue
> with me that doing _some_ things in Java is much easier than RPG,
> would you?

Yup.  I prefer Java for any type of *ML formatting, from XML to HTML.  I in
fact prefer Java for most instances of modeling static systems.  I dislike
Java for modeling variable systems such as business rules, because of the
two coding methodologies that support such systems, one performs miserably
and the other is easier to program using a procedural language that actually
supports procedures, such as ILE RPG.


> I also dislike it when people post that something works just
> fine when in all actuality they haven't done specific testing.  I felt I
> relayed my message so that people would understand that the
> "fastness" of my
> Java programs was completely related to my perception.

And I was reinforcing that it was indeed your perception, that's all
<smile>.

Joe