RE: Autoscaling?

[Chris Bipes <chris.bipes@xxxxxxxxxxxxxxx>]

I have not heard of an OS feature to handle this.  What I do is have my
servers look at the time stamp of when an entry was placed on the queue.  If
over xx time, submit another job to process.  When a job starts, you can
increment a data area.  When a job times out waiting for a data queue entry,
the job ends and decrements the data area.  You also create a control data
area with the time outs, min and max number of jobs.  At subsystem startup,
zero the counter data area.

You can also use a job to peek at the data area and see if there are entries
waiting and how long they have been waiting for.  This monitor job then
keeps count or uses system job api's to see how many data queue server jobs
are running and decides to start more or end excessive jobs.

Chris Bipes

-----Original Message-----
From: Metz, Zak

My company is looking into converting some COBOL to be threadsafe, but
we quickly discovered that this won't really work the way they want on
the 400 (we develop software for many platforms). I would like to
suggest that they instead use a system whereby requests are placed on a
data queue and jobs pick up those requests from there. I believe that
there is some OS/400 mechanism by which the number of jobs waiting on
the data queue will dynamically adjust based on workload (like the HTTP
jobs do). Could someone be so kind as to share what that sort of
processing is called? I will be happy to research it from there, but
without knowing what it's called I'm not getting far!