Re: Trigger Question

[Scott Klement <midrange-l@xxxxxxxxxxxxxxxx>]

> A discussion of Triggers the other day got a little heated when one member
> of the discussion said that IBM recommends that Trigger programs should not
> be designed to SBMJOB jobs but rather should do all of the processing within
> the bounds of the trigger program itself.

A trigger is fired every time the "event" occurs.  Let's say that every 
time a record is added to a file, you want to submit a job to do some 
processing.  This means:

a) Every time an application writes a record, a new job will be created.

b) If a program writes 10 records in rapid succession, 10 jobs will be 
submitted.  You'd better be careful to make sure those jobs can run 
concurrently and not conflict with one another.

c) If you copy a file with 10000 records onto the file with the trigger, 
you'll submit 10000 jobs, which might cause a big slowdown on the system 
as all 10000 are launched at once.


Even worse is when the event is on an update or delete, because you may 
have batch processes updating or deleting thousands and thousands of 
records, causing thousands and thousands of jobs to be submitted.

But, having said that... not everyone does that.  Sometimes people have 
files that aren't added to, updated or deleted rapidly.  Maybe in your 
business you have a file where one record is added one time per day.  In 
that situation, having a job submitted isn't a problem at all.

And that's the point I really want to make...   it all depends on your 
business and how it works.  YOU have to decide what's acceptable and 
what's not.

I personally think that data queues (or a similar means) are a much better 
idea than submitting a new job for every record, since they allow you to 
control how many jobs get submitted at any one point, and therefore keep 
things running smoothly.

Doing things within the bounds of the trigger might be good, too...  it 
really depends on what you have to do!