Page 1 of 1

Job Control or Job Sequence for Parallel Development

Posted: Tue Jul 28, 2009 6:37 am
by algfr
What do you suggest ? I'm used to Job controls since I have started initially with DS Server Edition. Is it better to switch to sequencers ?

Thanks

Posted: Tue Jul 28, 2009 6:51 am
by miwinter
When you say "for parallel development", does this mean:

a) development in Parallel (Enterprise) Edition of Datastage; or
b) parallel development of code (as in mutiple people working on the same thing at the same time)

Without this being clear, it's hard to gauge what you are asking really, though I do have some views on job sequences.

Posted: Tue Jul 28, 2009 7:14 am
by JRodriguez
Sequences are the way to go ....

Posted: Tue Jul 28, 2009 7:25 am
by miwinter
I differ slightly there. Sequences might be the way to go :)

Posted: Tue Jul 28, 2009 7:53 am
by algfr
Sorry if I was not clear. I mean that my elementary jobs are parallel jobs. i'm the only one developing

Posted: Tue Jul 28, 2009 7:58 am
by chulett
Parallel or Server has no bearing on this conversation as either solution is equally valid for both. If you have a robust "job control" solution in place then there may be no need for Sequence jobs, perhaps only leveraging them for any situation that your job control cannot handle, if any.

Posted: Tue Jul 28, 2009 8:01 am
by miwinter
OK thanks :)

Well, here's my two cents...

It really comes down to what level of complexity you need to achieve. Personally, I dislike job control and sequences provide much more flexibility. That said, it depends on your environment. If your scheduling tool allows complex logic for dependencies/rerun/handling etc, then it can often be better to code at a more granular level, avoiding sequences and having a job by job call from the scheduling. Anything in between creates a mix of the two which just confuses standards further.

Sequences also have issues of restartability. Whilst they are restartable from point of failure and restartable from the beginning (reset), they do not permit a partial rerun from a chosen point in the flow, so if you have bundled many jobs in a sequence, you may find yourself manually running parts of this instead. In scheduling, you might have found it easier to "rewind" to a chosen point and start again, in comparison to a sequence-driven design.

In the end, it's just a blend of design consideration along with the tools you have available to you, taking a wider view than the Datastage suite to include scheduling tools.

Posted: Tue Jul 28, 2009 8:21 am
by algfr
Great thanks

Posted: Tue Jul 28, 2009 4:36 pm
by ray.wurlod
All a sequence is is a GUI that generates a job control routine (which you can inspect in the job properties window). Sequences are preferred for anyone who is not an expert (and I use that term in its strongest sense) in the DataStage BASIC programming language and in using the DataStage API.