Re: JVM initialization memory concerns..?

[Pete Helgren <Pete@xxxxxxxxxx>]

Starting from the bottom and working backwards....

You could run a Java servlet through something like Tomcat or Webshere 
and use web services for the call (might be clunky).
Aaron Bartell uses a data queue in conjunction with a RPG/Java program 
in his RPGChartEngine (http://www.mowyourlawn.com/RPGChartEngine.html) 
so that he instantiates one JVM instance and accesses it repeatedly by 
queuing requests through the data queue.

I do believe that each invocation would result in a new JVM instance and 
yes, with multiple accesses my guess would be some performance impact.  
However, the JVM team at IBM has done some wonderful things with the JVM 
so they may have some under the covers magic that I am unaware of that 
minimizes the impact of JVM warm up.

Pete




Jonathan Keeney wrote:
> I have what feel like some very rookie questions about running Java apps on
> i...but please keep in mind that I'm very much a rookie. :)  Thanks in
> advance for any advice you all have to offer.  To give a little background,
> I have a set of Java classes that build a transaction object and pass it to
> an RMI service that calculates pricing information.  To populate the
> information in this transaction object, I've prototyped several methods in
> these classes and am calling them directly from a new RPG module I created.
> This module is intended to serve as a bridge between our legacy RPG code and
> the new calculation service.  It's functioning very well, but my concerns
> relate more to performance.  We're running V5R4 on a model 520, and we've
> run the upgrade so that we can run under Java 1.5 (5722JV1 option 7).
>
> - Am I right in thinking that a new JVM is initialized for each interactive
> job that will be calling this new module?  I've read that the JVM is
> integrated with i5/OS in a unique way that makes it function differently
> from JVMs in other OSes, but I'm not well-versed enough in these things to
> understand the repercussions of that.
> - If this is the case, is it reasonable to expect that our i5 would be
> overly taxed by having a large number (25? 30?) of these jobs running at
> once?
> - Is there any way to initialize one JVM and have all these sessions share
> it?  Would that even be beneficial if it were possible?
>
> Thanks again.
>