RE: JAVA400-L Digest V2 #56

I'd definitely expect that my web serving time would be higher than yours,
since I'm using the JVM and you're not.  The fact that your RPG time is
larger is primarily due to memory and processor.  I've got a 170-2285, which
has a 460 CPW, about 6 times faster.  I'm also running 3GB of RAM, ten times
what you have.  Of course, since IBM refuses to use a simple clockspeed,
it's kind of hard to judge.  We should run some vanilla benchmarks as well:
read a million records, perform a billion loops, that kind of stuff.

Personally, I'd expect that if we used the same architectural model on the
same machine, our RPG times should be similar and our web server times
should be somewhere in the area of 4-1 in your favor.  The interpreted
nature of the JVM would be enough of a problem to offset the
re-initialization of the CGI programs.

But I'm gratified to find that we're in the same ballpark.  You use 30
seconds, I use 27 seconds on a 6-times faster machine.  A straight
extrapolation would extend me to 162 seconds, 5.5 times as slow, which is
pretty close to my estimate above.  And remember, this is for pretty
brain-dead applications.  As the RPG programs go up in complexity and the
backend processing increases, that ratio shrinks.  I know I have to have
more muscle to run Java, but the fact that it's not THAT much more is pretty
nice.  And I'd still like to do some straight power-program testing to see
if the programs run faster.

More importantly, look at the gross numbers.  If a complex (8KB) page takes
290msec to process, that limits me to... 10,000 hits per hour.  A quarter
million a day.  On a model 170.  Not too shabby.  That means I can get a
little 12-way processor and crank that up to 5 million hits a day <grin>.

Also, I've got lots of debug code in my stuff (I generated over a megabyte
of STDOUT prints), and I haven't cranked up the optimizations yet.  That
might help.  But even if Java is four times as slow - the flexibility I have
in my design makes it worth it, I think.

Joe


> I mentioned earlier that I use an HTTP server plug-in (message server) to
> route requests to the RPG application server (GLCAMT).  The
> downloaded HTML
> is 5.8K per page, which is 76% larger than your 3.3K page.  My HTTP Server
> used only 10 seconds which is half of yours.  But my RPG program
> use nearly
> 20 seconds which is more than double that of yours.  The larger page size
> would account for some of the difference, but not all.  I'm wondering
> whether I need to fine-tune the API used by the RPG program?
>
> Also, I'm using a 170-2290 with 320 meg RAM with a CPW rating of apx. 70.
>
> What's your machine?

+---
| This is the JAVA/400 Mailing List!
| To submit a new message, send your mail to JAVA400-L@midrange.com.
| To subscribe to this list send email to JAVA400-L-SUB@midrange.com.
| To unsubscribe from this list send email to JAVA400-L-UNSUB@midrange.com.
| Questions should be directed to the list owner: joe@zappie.net
+---