RE: Single record access really required (was RE: Views and

> From: Dave Odom
> 
> What does IAAI stand for and how are they involved with performance
> issues?

Dave, the IAAI is the iSeries Application Architecture Initiative; its
website is http://forums.plutabrothers.com/IAAI.  It came into being
after a number of discussions about performance and the like here on the
mailing lists and in the iSeries community in general.  The final straw
came recently when someone insisted that JDBC is as fast as native I/O.

The problem comes when you compare apples to oranges, as you say.  It's
my belief that in the traditional record-at-a-time processing we do in
standard OLTP applications, that nothing beats RPG and native I/O.
We've heard tell from people who insist that Java and EJB with
commitment control will actually outperform native I/O, and of course
the SQL discussion Charles and I just had outlines the opinions there.

The problem is that everything is anecdotal.  And because of that,
whoever is selling something can pretty much point to a specific
circumstance and say "See? My solution runs better!"  The problem is
that they then generalize a single circumstance to all applications, and
thus we have these coantradictions.

So I said the heck with it and created a website.  The site is going to
be dedicated to finding the real answers to things.  We're going to
create real test beds and measure performance.  We're going to compare
technologies head to head for things like security.  We're going to
identify different architectures, and maybe, just maybe, we'll create a
large enough repository of information so that people can come in with
normal business requirements and be able to ascertain which
architectures are best for them, without having to wade through hype.

For example, I think I've made it clear that at least in some
circumstances, native I/O FAR outperforms SQL.  On the other hand, with
the help of folks like Charles we may uncover some new paradigms in
which SQL is a better fit.  I suspect the same will be found when we
compare CGI to JSP or Java to RPG.  But, unlike the anecdotal evidence
found on mailing lists (and in vendor sales pitches), we'll have real
numbers to back it up.

It's an ambitious project.  There's a lot to do, and right now only a
few of us are finding the time to post.  I plan to start a few
conversations at least in part based on discussions we've had here and
hopefully that will help get the ball rolling.  If nothing else, Charles
and I will have a place to share our source code <grin>.

So, please, come by and help.  I'd like this to be a joint effort by the
entire community.  Brainstorming will be allowed, but at the end, I hope
we can come up with real, hard numbers and publish them on the website
for the world to see.

Joe