Re: CGI vs JSP...

[Mel Rothman <mel@xxxxxxxxxxxxxx>]

I assume you ran the CGIDEV2 and RPGSp programs repetitively.

In the code you posted today, the CGIDEV2 program runs in a *NEW activation group, which 
is a real performance killer.

Try running it in a named activation group. The first execution will be relatively slow as 
the program reads and processes the template. Subsequent calls will use the already 
processed data and run much faster.

Also, you have debugging on in all its glory.

Try changing
  h dftactgrp(*no) actgrp(*new)
to
  h dftactgrp(*no) actgrp('SOMENAME')

Try adding the following statement at the top of the program:
  setnodebug(*on);

These changes should greatly improve CGIDEV2 performance, significantly closing the gap 
between RPGsp and CGIDEV2 performance.

In case others hadn't noticed it, Seth's signature indicates he works for Profound Logic 
Software, Inc., the company that sells RPGsp.

For those who don't already know it, I authored CGIDEV and CGIDEV2 during my tenure in 
IBM, which ended in 2002. I continued supporting and enhancing CGIDEV2 until the middle of 
last year when Giovanni Perotti left IBM and IBM Rochester's CTC organization became the 
its sole owner and maintainer.

I surmise that Seth is proud of his company's products as I am with CGIDEV2 and none of 
these statements are intended to be disparaging in any way.

For people who want free CGI tools, CGIDEV2 is a great alternative.  For those willing to 
pay for commercial products, apparently RPGsp is an attractive offering (I have no direct 
experience with it).

Mel Rothman
Mel Rothman, Inc.

Seth Newton wrote:
> Brad,
>
> RPGsp is smart enough to detect when only the HTML is changed, in which case
> it saves off the HTML, and does not recompile the RPG code.
>
> I ran some performance tests on RPGsp vs. CGIDEV2 and the results were
> astounding.
>
> My example displayed 1000 rows of dynamic data.  Each row had 30
> substitution variables.  I basically used the same HTML and RPG code in both
> tools.  I displayed a timestamp at the top and bottom of the page.
>
> RPGsp consistently displayed the information in 0.2 to 0.3 seconds.
>
> CGIDEV2 was consistently between 8 and 10 seconds, about 30 times slower.
>
> By the way, I ran this on a 520 500 CPW machine, and I did turn debugging
> off in CGIDEV2.  I am convinced now more than ever that RPG CGI is faster
> than Java and any studies that show otherwise are based on specific tools
> that may not implement CGI efficiently.
>
> Seth Newton
> snewton@xxxxxxxxxxxxxxxxx
> Profound Logic Software, Inc.
> Toll-Free: (877) 224-7768 x115
> Fax: (603) 849-7757
>
> RPGsp - iSeries Web Development has never been this easy!
> Watch video demos: http://www.profoundlogic.com/video_demos/
>
>
> ------ Original Message ------
>
> So if the template changes, that means a recompile though, right?  Now, I
> know that a recompile shouldn't be a "big deal", but I know one of the nice
> things about CGIDEV2 or other packages is that you don't have to recompile,
> just change the template.  Sometimes that feature outweighs the milliseconds
> one may save.  I used to say "who cares if I recompile", but once I found I
> didn't have to, it was something I almost couldn't live without.
>
> But, I would be curious as to some speed tests as well.
>  Theoretically RPGSPs can't be faster than raw CGI unless they use something
> else in the background.  
>
> The only reason I ask is, with my tests, on newer machines I haven't seen a
> noticable difference between CGIDEV2, eRPG SDK or raw RPG-CGI.  The
> bottleneck is the HTTP server (or the QtmhWrStout API) not being able to
> spit out the data fast enough, and/or the network.
>
> AS an example, I ran some tests on eRPG SDK vs CGIDEV2 on a VLP machine from
> IBM (a 570 partition I think).  Well, I couldn't get the execution of either
> to be more than a second until I went up to 50,000 iterations.  Even then
> while the app took a second to run, it probably took my browser 30 seconds
> to load the data.
>
> Brad