Designing Sequnece job

rafik2k · Post by **rafik2k** » Tue Apr 17, 2007 2:35 pm

I am in process of creating the sequence,
Basically I have two questions:

1) While designing sequence job, I am unable to open property of Job_Activity stage where as I am able to open property of other stage like Terminator_Activity stage etc.
Could you let me know is there anything that I need set in administrator or in any other place? or what is the issue?

2) As per my understanding I can design sequence job like following two ways

a)Job_Activity1----->Job_Activity2----->Job_Activity3----->more Job_Activity ( sequenctial flow)

b) Job_Activity1---------->Job_Activity2 (parallel flow)
|
|
|
|
Job_Activity 3

Can you plz let me know in term of permormance which above two will be executed faster?

or what is the best approach for designing sequence job.

Plz note that I am running server job i in Windows environment with 4 cpu

Any help//suggestion will be highly appreciated.

Thanks in advance
Regads,
Rafiq

ray.wurlod · Post by **ray.wurlod** » Tue Apr 17, 2007 2:50 pm

Almost certainly you need a patch to address the inability to open Job activity in job sequence, and my guess is the XPSP2 patch. Contact your support provider.

Your second question is irrelevant. The links in a job sequence indicate dependency. In your (a) job every activity is dependent upon all previous activities, in your (b) job this is not the case (for example JA3 is not dependent upon JA2). The designs are functionally different.

Design what you need, in terms of dependencies. The fastest design, presumably, will be the design with no links (no dependencies) at all, so that all activities will run simultaneously, up to the point where the hardware is overloaded.

rafik2k · Post by **rafik2k** » Tue Apr 17, 2007 3:27 pm

ray.wurlod wrote:Almost certainly you need a patch to address the inability to open Job activity in job sequence, and my guess is the XPSP2 patch. Contact your support provider.

Your second question is irrelevant. ...

Thanks Ray.
Though I can't see answer of second part query. One day I will see that also.

Regards,
Rafiq

us1aslam1us · Post by **us1aslam1us** » Tue Apr 17, 2007 4:24 pm

For your second query, It all depends on the dependencies between the jobs.
If your Job_Activity3 is independent of Job_Activity then you can run them in parallel and so on....

rafik2k · Post by **rafik2k** » Tue Apr 17, 2007 4:31 pm

us1aslam1us wrote:For your second query, It all depends on the dependencies between the jobs.
If your Job_Activity3 is independent of Job_Activity then you can run them in parallel and so on....

In my case all are independent jobs.
I am wondering , how is it possible to run job in parallel in server job??

ray.wurlod · Post by **ray.wurlod** » Tue Apr 17, 2007 6:08 pm

How is it possible to run job in parallel in server job??

Think about what you're asking. You want to run one thing in parallel? There are many options, but what are you trying to achieve?
A job sequence can run any mix of server and parallel jobs simultaneously.
A server job can execute job control routine to run multiple jobs, again any mix of server or parallel job types.
Parallel independent streams can be designed into server jobs, and will generate multiple processes.
Use of intermediate passive stages will generate multiple, though consecutive, processes.
Use of inter-process row buffering, IPC, Link Partitioner and Link Collector stages in server jobs will generate multiple processes.

What was it you wanted to do again?

rafik2k · Post by **rafik2k** » Wed Apr 18, 2007 1:04 pm

Thanks Ray,
Actually I wanted to know about parallelism of server job.

Correct me if i am wrong, now as per my understanding,if we have multi-processor system, then we can utilize the advantages of parallelism for server job.
For single processor system, we can't get this advantage,because behind the scence utlimately processing will be done sequentially, though we can have mulitple instance of a process(time sharing concept).

Regards,
Rafiq

ray.wurlod · Post by **ray.wurlod** » Wed Apr 18, 2007 3:41 pm

Not true. DataStage job will generate processes. You can run as many processes as the resources (CPU, memory, I/O bandwidth, etc.) will support. If you choose to use some of your resources running parallel (simultaneous) instances of things, that's your choice, even with only one CPU. If each process requires 19% CPU, then you should be able to run five of these if nothing else is happening.