Re: Embedded SQL - performance question

On 5/27/2011 3:22 PM, Alan Campin wrote:

I am just not sure about all this SQL performance arguments. There is a
difference mostly occurring I think because IBM continues to use a call
model instead of ILE call model but I wonder how much difference it really
makes.

We have about 70 trigger service programs in use daily written 100% in
ILE/RPG using 100% SQL for I/O being fired millions of times everyday but
our users have not reported any difference in performance since we turned
then on.

The performance issue is in the eye of the beholder. For some shops, it's the amount of time for interactive updates. For others, it's the length of the nightly batch runs. For still others, it's locks and other conflicts. That's why I don't usually talk about performance in terms of system usage, just in raw numbers. I think it's always good to know how different approaches compare; it can help to make design decisions. Of course, if you have more CPU and disk performance than you can possibly use, then the discussion is moot and any architecture will do. But I find that it never hurts to pay a little attention to the performance - it's a lot easy to design certain techniques in up front than to engineer them in later.

Joe