Calling Basic Routine in Parallel Job

[Biggles]

I am attempting to call a Basic routine in a parallel job - through the parallel transform it doesnt seem possible - I read here that it is possible to use aserver transform in a parallel job and therefore call the server routine - but that was a 7x post. Has anyone managed this in 8x ? Doesnt seem to want to fly.....

[ray.wurlod]

Welcome aboard. :D

Parallel routines (callable from parallel Transformer stage) must be written in C++. This stage type does not support DataStage BASIC routines. In any version of DataStage.

If you want to call a BASIC routine you need a BASIC Transformer stage. But this, too, has its downsides (Search for more information).

[Rubu]

I guess, correct me if I am wrong, a Basic Transformer can only be used in SMP servers.

[ray.wurlod]

BASIC Transformer stage can be used in MPP environments, but only in a node pool that contains only the node on which the DataStage server is installed (usually but not necessarily the conductor node).

[Biggles]

I guess when you decide to use a basic routine you lose the advantage of parrallelism so within a sequence I am using a mixture of Server and Parrallel jobs - server if I want to call a basic routine, otherwise parrallel. I am not loading an enormous amount of data so the performance difference is minimal.

We are reluctant to start building routines in c++ for maintainability reasons, lack of experience within the company etc....

[ray.wurlod]

Job Sequences themselves do not have any innate parallelism, and BASIC expressions (and therefore routines) can be used throughout.

Well-designed server jobs can process small to medium volumes of data faster than parallel jobs, particularly if the latter have not been well-designed, but also because of the startup overheads incurred with parallel jobs.

[DS_FocusGroup]

How would you define small to medium size data ? 5-10M ? 10-20M? 20M-30M ?

[ray.wurlod]

Up to 1GB to 2GB is still medium, depending on the hardware.