Polycenter scheduler alternatives

From: Guy Dallaire <dallaire_at_total.net>
Date: Fri, 14 Mar 1997 13:52:13

You probably know that DEC dumped it's polycenter products and that CA are
taking it over. Unfortunatly, they do not provide big support for
polycenter scheduler and are instead pushing their own UNICENTER product
instead.

We are presently using polycenter scheduler 3.0 under digital UNIX 3.2D-1
to run our batch jobs, but we fear that CA will completly drop Polycenter
scheduler. Anyway, polycenter scheduler is not supported under DU 4.x and
we will prabably move to DU 4 one of these days.

We have two options:

1) Move to UNICENTER TNG.
2) Create our own "pseudo" scheduler.

Option 1 will probably require us to rewrite/redesign most of our batch
jobs in order for them to function in the new scheduler environment. Not to
mention that my previous experiences with CA's product were not necessarily
funny (Bugs, "features", etc...) So we are wondering if solution 2 would
not be a better way for us to do it. Not to mention that Polycenter
Scheduler was a real pain to manage in a DecSafe environment.

Option 2:

We will need BASIC scheduling services. Crontab is out of the question
because our batch system has to interface with our production database and
we don't want too much straps, pulleys and/or band-aids. I would like to
know if any of you is aware of a GOOD, STABLE, EASY TO USE batch queue
system under DU. Something like VMS queues would be fine (We have to be
able to open or close queues based on a certain period of time, hold jobs,
release jobs, etc...)

The scheduling system have not to be centralized on a single server. We can
have each server run its own batch processing system, but if one of the
servers goes down, we would like to be able to run the jobs from one server
on another server. (all the jobs, programs, etc..) will be failed over to
the surviving machine.

Most jobs that we run are made of separate programs or scripts, called by
ksh scripts. There is a level of dependency between the steps a job is made
of. There is also dependency between jobs themselves. We MUST have a way to
determine if a job: FAILED or COMPLETED SUCCESFULLY. Automatic requeuing of
jobs would also be a plus.

Any help/suggestions would be appreciated.



Guy Dallaire
dallaire_at_total.net

"God only knows if god exists"
Received on Fri Mar 14 1997 - 20:07:04 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:36 NZDT